Vulnerabilities found in a decades-old open-source web technology almost “broke the internet” recently, said Zack Whittaker in TechCrunch. The existence of bugs in PHP, a 31-year-old programming language used by roughly 70% of all websites and web apps worldwide, had been known for years “but weren’t considered exploitable.” That changed with AI. Security researchers using frontier OpenAI models reported that they “found a way” to weaponize those flaws, raising the risk that hackers could “break into any unpatched system on the internet without needing a password or any other credential.” The researchers notified the PHP Foundation, which released patches to fix the bugs—but companies still must race to update their websites before any hackers break into them. Matthew Prince, the chief executive of cybersecurity giant Cloudflare, said there have been “an exponential number of attempts to exploit the bugs following their disclosure.” It’s likely we’re not yet “out of the danger zone.”

“This is the cyber equivalent of an unexploded bomb,” said Zoe Kleinman in BBC.com. The weaknesses in PHP have been “quietly resting there for years”—until AI arrived and “the clock started ticking.” So far, the bomb “remains intact,” with no reported incidents, though it may be a while before the full extent of the damage is clear. Web developers will need “thoughts, prayers, and maybe a strong coffee” to keep their systems safe. The welfare of the entire internet comes down to one man, said Charlie Warzel in The Atlantic. Ondrej Sury, who lives in the Czech Republic, is responsible for maintaining and automating the patches for PHP. He is an “unsung hero,” and a reminder of how much of the internet’s infrastructure “is maintained by small crews of volunteers”—against an army of bots.

That’s why it will take “more than a handful of exhausted developers to keep us safe,” said Chris Stokel-Walker in The Guardian. The PHP Foundation, a nonprofit group of users, says developers identified the issue in 2024, but it was “too expensive and too tricky to fix.” There’s an easy solution here. Big Tech companies make billions of dollars from open-source software they essentially “get for nothing.” They could easily allocate resources to keep their platforms’ foundations from collapsing. Unless more companies act, this is only the beginning, said Matteo Wong in The Atlantic. “More websites will be defaced, more personal information will be stolen, and ransomware will proliferate” as AI-powered hacking matures. Organizations that have been “lulled into a false sense of security”— relying on older software and the idea that only a few skilled hackers could get access—should be “waking up from a trance” now that anyone can deploy thousands of AI agents to sniff out vulnerabilities and automate attacks. By next year, PHP-ocalypse may be “remembered as normal, or even tame.”

Investors withdrew $3.6 billion (£2.7 billion) of assets from the cryptocurrency exchange Binance when it reopened for business after what is thought to be the biggest financial theft in history.

Hackers stole around $6.3 billion on Thursday, though Binance claims to have “replenished” around 90% of that into customer accounts.

It took “no more than 30 minutes” to move the “eye-watering” sum, said the Financial Times. The funds may prove “impossible to recover”, said CoinDesk, renewing the “debate over the safeguards exchanges need to prevent a security breach from becoming a financial catastrophe”.

How did it happen?

Crypto exchanges like Binance, which has more than 300 million registered users worldwide, are especially “tempting targets” for hackers, said The Wall Street Journal. While traditional stock exchanges don’t hold the securities of their investors, and only facilitate trades, crypto exchanges often store their customers’ cryptocurrencies. Hackers can therefore directly access customers’ funds from the exchanges’ crypto wallets.

Binance co-CEO Richard Teng said on X that the security breach could have originated via a “compromised” laptop belonging to one of the company’s engineers. However, this would not have been a “single ‘one-click’ action”, but a “more complex attack process”.

Who was behind it?

Crypto analysis firm TRM and other researchers have attributed the attack to a North Korean hacking unit TraderTraitor, based on “robust on and off-chain intelligence”. This has yet to be confirmed.

“If North Korea is proven to be the culprit, it would be little surprise to US law enforcement officials and private experts,” said CNN. TraderTraitor has been “behind many of the largest crypto hacks in recent memory”, said WSJ, including the previous record theft of $1.5 billion last year from Bybit, another major cryptocurrency exchange.

What will happen next?

Binance said on X it has frozen more than $3 billion (£2.2 billion) – around half – of the stolen crypto and promised to cover all money lost in the hack from existing funds.

Freezing the crypto means the hackers “cannot move, transfer or cash out” their takings, essentially “trapping the tokens inside the wallet before hackers can convert them into other cryptocurrencies or launder them”, said WSJ.

Recovering the losses may prove more problematic. This requires a “separate legal process” involving court orders or a seizure warrant, Ari Redbord, from TRM, told CNBC. Recovering Bitcoin in particular – which accounts for around a quarter of the haul – is “more challenging because they have no issuer that can directly freeze the assets”, said Eric Jardine, from Chainalysis. “There is no standard recovery timeline. A technical freeze may happen quickly, while returning funds can take months or years.”

With many American tech companies looking to expand their footholds across the world, a large number have planted their flags in Aragon, a diverse region of northeast Spain. While Aragon is one of Spain’s least-populated areas, its tech boom is quickly becoming consequential.

Why Aragon?

Many U.S. conglomerates are moving there because of Aragon’s sparsity. The area is a “Big Tech company’s dream,” as “vast, uninterrupted stretches of the region are empty,” said Bloomberg.

Aragon’s capital, Zaragoza, is Spain’s fourth most populous city, but beyond that the “landscape opens out into broad countryside where most municipalities have fewer than 26 residents per square mile,” said Bloomberg. This large territory has made Aragon one of the “fastest-growing data center hubs, with investments announced by Amazon, Microsoft and others surpassing $94.4 billion.”

Amazon is one of the biggest investors and “promises to spend almost $40 billion to build multiple vast complexes — one alone as large as 600 American football fields,” said NPR. Amazon and others like it believe Aragon “could serve as a model for a European Union seeking to triple its data center capacity within seven years and narrow the AI infrastructure gap with the U.S. and China,” said Reuters. Unlike other areas in Spain, Aragon’s government is also largely friendly to data centers.

What does this mean for the region?

Many officials “speak boldly of the jobs and long-term prosperity that await once Aragon becomes ‘the Virginia of Europe,’” said Bloomberg, referring to Loudoun County, Virginia, which has the world’s highest concentration of data centers. The Aragon government “expects data centers to create tens of thousands of jobs and says local companies are already feeling the positive impact.”

Still, questions about long-term viability remain. Due to fears about the AI bubble bursting, some investors have “cast doubt on the billions of dollars that tech companies are pouring into infrastructure upgrades and partnerships with AI startups,” said The Wall Street Journal. The rapid buildup of tech has also “left many in Aragon feeling excluded from decisions that will profoundly impact their region and lives,” said NPR.

“We are going to open the window and see nothing,” said Silvia Baselga, who lives near a data center construction site, to NPR. It’s “going to be like living beside a prison.”

AI is causing “a crisis of trust” in publishing, said Vauhini Vara in The Atlantic. Earlier this month, Zimbabwean writer Simbarashe Steyn Kundizeza won the prestigious Caine Prize for African Writing—worth $13,000—for a short story that “has all the personality of a chatbot.” That may not be coincidental: Pangram, an artificial-intelligence-powered AI detection tool, flagged “almost the entire story as AI-generated.” Kundizeza is not the first award recipient to be flagged by Pangram; it deemed one of the winning stories in this year’s Commonwealth Short Story Prize to have been entirely authored by AI. Kundizeza says his story, “The Blue Fighter,” is “entirely my work.” Pangram is reputed to be “highly accurate”—the company claims its tool misidentifies AI use in just 1 out of 24,000 cases—but it’s not infallible. So who should we trust? The author or Pangram?

This is exactly the “chaos” that publishers feared, said Anna Silman in The Wall Street Journal. Publishing houses have tried to crack down on AI usage by authors, inserting strict clauses into book deals and leaning on Pangram to vet manuscripts. But “a new AI scandal engulfing” the industry seemingly every month is “forcing a reckoning over the nature of authorship.” The Big Five publishing houses—Simon & Schuster, Penguin Random House, Hachette, HarperCollins, and Macmillan—have been reluctant to “make sweeping pronouncements about” the technology, possibly because they’re waiting to see if people care whether the books they’re reading are AI-generated or AI-assisted, “especially in genres like fantasy and sci-fi.” Meanwhile, many writers feel that a contract clause “that relies on the honor system lacks the teeth necessary to combat” intellectual dishonesty.

I write a lot, and I use AI, and I’m worried that “I’ve started to internalize” how a chatbot thinks, said Alex Kirshner in Slate. I’ve begun to “self-police against this,” modifying my drafts “to sound less like AI.” I also put my work into Pangram “before it appears anywhere else” to ensure that “the model concurs that I wrote it myself.” It feels dystopian—I’ve had to “rewrite sentences of my own work in hopes of satisfying this new integrity god.”

Still, we need the “AI slop detectors,” said Brian X. Chen in The New York Times. The tools do more than pick out obvious indicators of AI authorship, such as the excessive use of em dashes and bullet points. They’re trained to understand how AI models make decisions on word selection and phrasing. Pangram, which costs $20 a month, is incredibly “useful for sniffing out annoyances like AI-written email scams and phony online reviews.” Yes, the technology “still has a long way to go.” But “at the rate that slop is proliferating on the web,” it’s an “empowering” tool for people who still want to read the human-written word.

Economists have spent much of 2026 watching to see if the artificial intelligence bubble will burst, much like the boom-and-burn cycle of dot-com pioneer companies in the late 1990s. But another possibility looms. The wild enthusiasm for AI could settle into a sensible soft landing that spares the world a hard economic crash.

‘Reassessment under way’

Wall Street continues to believe that AI will “transform markets and the economy,” said Axios. But investors are also “pulling back funding” and looking for “evidence that their investments will actually pay off.” The markets that had a “blind enthusiasm for all things AI” are now beginning to “differentiate between players, business models and financing approaches," said JPMorgan’s David Lebovitz to the outlet. That does not have to be scary. Call it a “rational exuberance phase,” said Axios.

The recent decision by Australian data center company Firmus to pull back from a planned initial public offering is “triggering the usual angst” about the AI bubble, said Patrick Commins at The Guardian. The growing skittishness by traders “doesn’t mean the whole thing is a disaster,” said NAB economist Sally Auld to the outlet. But there is a “reassessment under way” of AI’s anticipated returns.

The damage from a popped AI bubble would be “severe but not permanent,” said Ruchir Sharma at Foreign Affairs. It will likely follow the cycle of previous technological revolutions, with the rise of artificial intelligence “moving forward even as the speculative frenzy around it fades.”

‘Double whammy’

Fears remain. The “untold billions poured into AI” have driven the economy, but the “boom could quickly invert” if profits are elusive, said Futurism. The fallout could ironically result in job losses and end up “accelerating the replacement of workers with AI,” said Greg Rosalsky at NPR’s Planet Money.

One possible solution to that “double whammy” scenario is a tax on AI and robots that would make it “economical to continue employing humans” while funding the safety net, said Rosalsky. But there is a less-frightening path forward, as well. A downturn could “leave businesses with less money to invest in AI,” making them reliant on human workers.

The UK is facing a greater threat “than we have, certainly for most, if not all, of my lifetime”, Wes Streeting told the BBC.

The defence secretary was referring to the vulnerabilities discovered in PHP, a programming language used in up to 70% of websites. Since then “millions of attacks” have been launched in attempts to exploit the flaw, said The Times, and a “patch” has been rolled out to help sites deal with the issue. The government announced 32 of its services were offline, though there is no indication of a breach.

For many, this episode has highlighted the dependence on, and vulnerability of, internet use. With AI growing ever more powerful – and likely being used by hackers as well as those researchers who found the PHP vulnerabilities – this could be just the beginning of an era of disruption.

What did the commentators say?

In the last few days, the internet has been “tested in a way I have not seen in my – or its – lifetime”, said Andrew Griffin in The Independent. Usually global catastrophes “capture our attention and excite our energy”, whereas the “gradual weakening of things we depend on engenders boredom”. But we need to be aware that AI does threaten the world, just not in a traditional way. It does so in a “way so dull and disengaging that we don’t even realise”.

“The future is here” and our digital infrastructure is “looking very fragile”, said Chris Stokel-Walker in The Guardian. “The solution is the same one we keep talking about, but rarely actually do anything about.” Large tech companies, who make “billions” from software they are obtaining for free from open-source data, “should contribute properly to maintaining it”. Unless we start “properly valuing those who keep our internet safe”, the internet will break, again.

There have been warnings of this crisis for the “better part of a year”, said Matteo Wong in The Atlantic. In the last few weeks, there have been major cyberattacks on “governments, data centres, and hospitals in Ukraine, Russia, Argentina, Bangladesh and Indonesia”.

“Bulking up” IT departments and using passkeys and two-factor authentications are still “important”, but the “most urgent and practical advice being given now is to retreat: prepare to simply go offline”. There is little time now to prepare for the “deluge” of AI-based hacking. In fact, “the dam may have already broken”.

The PHP “patch” does not spell the end of these “dramatic” issues, said an editorial in The Times. “Millions” of websites will still need to be updated, despite the fix, and even then, hackers could use the opportunity to “illuminate their route into systems that may still be sitting unprotected”.

Government preparedness should be an absolute “political priority”. Pilot projects on cyber defence and the AI Safety Institute represent “sensible” first steps, but “more must be done” in the form of public awareness and creating “decisive state emergency levers” in times of crisis. We should not panic, but the coming weeks will “no doubt reveal continuing threats, especially to smaller bodies with less stringent security”.

What next?

“There’s a race on” between attackers and the tech companies who are rolling out the “patch”, said Rowland Manthorpe on Sky News. “The trouble is, this patch doesn’t cover all PHP sites.” Based on data gathered from around 2.6 million services, about 75% don’t run newer versions of PHP, and so it will be “much harder” for them to implement the fixes.

In the long term, AI “could be amazing” for developing cyber defences, but for now it seems the “attackers might have the advantage”. So “we could be in for a bumpy few years”.

Togo has asked its citizens to contribute texts, recordings and translations to “develop AI models that address the linguistic diversity of the country”, said AI for Developing Countries Forum.

More than just a “regional experiment”, it is a “blueprint for how developing nations can use technology to assert their cultural sovereignty and economic relevance in the digital age”.

Africa has around 2,000 languages but very few are represented comprehensively in AI models, said academics on The Conversation. Failing to be a part of the tech revolution not only “marginalises millions of people” but also “increases the technology divide”.

98% of African languages ‘unsupported’

AI’s inability to handle most African languages comes down to the “gap between the number of real-world speakers and a language’s digital footprint,” said African Business.

According to research published in the Proceedings of Machine Learning Research, across six large language models (LLMs), eight small language models and six specialised small language models, only 41 African languages were represented. Only four – Amharic, Swahili, Afrikaans and Malagasy – were “always handled”. “Over 98% of African languages went unsupported.”

According to OpenAI, the number of ChatGPT users in Africa is “growing faster than in any other region, said The Economist. Yet, despite the surge in popularity, even the best LLMs “falter in African languages, typically scoring 10-30 percentage points lower on benchmarking tests than they do in English”. The models mainly struggle with numerous dialects and “code-switching”, which is when speakers use different languages at the same time.

“The root problem is lack of data.” Much of the training for “smaller” languages is taken from Bible verses. High-quality audio fit for speech recognition is also “scarce”, so researchers are “recording how people talk” to build the data supply.

It’s not just a lack of text and speech that is the problem, said the academics on The Conversation, it’s also the infrastructure that supports it. “Basic tools” like dictionaries, terminologies and glossaries are missing. Then there are inadequacies in African language keyboards, fonts, spell-checkers and tokenisers – which “break text into smaller pieces so a language model can understand it”. All of these impact orthographic variation – how the same words are spelled differently across regions – tone marking and “rich dialect diversity”.

‘Global North cultural assumptions’

The texts that dominate LLMs “reflect Global North cultural assumptions, about individualism, competition and about globalisation”, said University World News. By definition, therefore, this is a continuation of the “long history of epistemic marginalisation” of the Global South.

The lack of African languages in LLMs has become a “major concern for the continent’s policymakers and tech leaders”, said Semafor. The World Bank estimates that Africa has only 0.6% of global data-centre capacity, despite having 18% of the world’s population.

Africa’s “main link” to the AI revolution has been through minerals such as cobalt and copper, but the “bigger prize” is through productivity, or being able to harness the end product of AI. Currently, however, there isn’t the infrastructure or energy to achieve this, meaning that AI is “unlikely to add meaningfully to growth”.

African AI companies need to break beyond being “suppliers to the AI revolution or just consumers of finished AI products”. Their continent needs to be the place where the “revolution delivers its biggest gains”.

Building AI models from scratch is “expensive” and uses “limited resources”. By adopting and adapting existing models, the continent’s outputs can evolve more quickly. And Togo’s initiative suggests that the future of AI in Africa “may hinge less on who builds the biggest model than on who makes the technology useful to the most people”.

For the first time in years, applications are flowing in to the Internet Corporation for Assigned Names and Numbers (ICANN), the organization that controls web addresses, for new top-level domains. These proposed suffixes at the end of URLs will shape the internet’s next era.

What did the applications reveal?

ICANN has received 1,615 applications from 481 applicants so far, according to its Reveal Day announcement earlier this month. It is the “biggest shake-up” of internet addresses in “more than a decade” and could “result in you typing .fart to reach a website,” said The Times. While some domains are “quirky,” others show how “AI companies are seeking to grab domains like .agi, .superintelligence and .chatgpt.”

Ten different companies including Meta and OpenAI applied for the .agent domain, seven for .agi, and six including OpenAI for .asi, “attempting to get in on” Trump’s push to switch from artificial intelligence to super intelligence, said The Verge. OpenAI applied for 15 top-level domains, or TLDs, while Meta filed for 21.

This is the first time ICANN has accepted applications in over a decade. The nonprofit “first sought new top-level domain applicants in 2012,” said Politico, resulting in .app, .web and .pizza. Each application costs $227,000 and is “subject to, among other processes, a community input period that runs from November to March.”

Those awarded a domain “become registries, gaining control over the suffix,” said The Times. They can keep these for themselves or “sell the registrations under it to third parties.”

Who benefits?

The domain applications show that organizations worldwide are “seeking new ways to build trusted online identities, serve their communities and connect with internet users,” said ICANN President and CEO Kurtis Lindqvist in a press release. But not everyone is enthusiastic.

Internet safety expert John Carr, who has attended ICANN meetings for a decade, is critical of the organization’s process. It’s a “conspiracy against the public good,” he said to The Times. Under its rules, ICANN is supposed to check the “real-world identities” of the people buying the domain names, but “they don’t.” These new domains are just “another way for them to make money.”

The data center boom has “entered a new, contradictory phase,” said Brian Martucci in Utility Dive. Construction activity on the massive warehouses filled with AI servers “remains robust.” But several billion-dollar projects are facing “a host of challenges,” including worker shortages, a lack of gas turbines, grid constraints, permitting delays, and widespread community opposition. The problems came to a head recently in New Mexico, when Oracle issued an unusual force majeure notice—normally reserved for insurance against “acts of God”—to “protect itself financially from the delay or cancellation” of its so-called Project Jupiter, after state regulators “blocked a gas pipeline proposed to serve it.” Google has also recently had to halt work on two data centers in Finland after a local outcry spurred an environmental review, while a Microsoft project in Michigan “faces an uncertain future amid local permitting delays.” Goldman Sachs estimates only 50% of planned data-center capacity may actually come to fruition in the near term.

That means the AI boom may soon hit a bottleneck, said Asa Fitch in The Wall Street Journal. Tech giants need “lots of data centers and electricity” to fulfill their ambitions, but the “delays are biting.” The amount of electricity demanded by data centers globally in the next two years “is enough to power around 118 million homes.” It’s curious that “chipmakers aren’t yet blaring their sirens” about the capacity squeeze. Possibly, that’s because they can’t even accurately tell “how much is being built”—and how much is being idled.

This “cornucopia of holdups” should be a wake-up call for investors, said Chris Bryant in Bloomberg. AI companies are spending billions “to secure precious supplies of computer memory” from chip makers like Nvidia and AMD. But if they can’t build data centers, they could end up “with too many chips and not enough operable server farms.” Almost half of the U.S. data-center pipeline is being built by first-time developers who have jumped into the gold rush, making big promises they might not deliver. It’s a “high-stakes game of musical chairs,” where losers will be left holding chips “without a seat.”

“Why assume the data center has reached its final physical form?” asked Todd G. Buchholz in The Washington Post. Historically, as technology has advanced, it’s gotten smaller. In 1946, a computer “took up the size of a house”; now it “fits in the palm of one’s hand.” Likewise, we can safely assume that the land, electricity, and equipment needed to produce a unit of intelligence will keep dropping. “That’s why communities should welcome new investment but insist that developers pay the costs of power, water, and grid upgrades.” The benefits from the technology will last for years, but the size of the physical footprint will not.

A new United Nations report is sounding the alarm on another potential risk of our constantly evolving world: technology that could allow governments to “read minds” via brainwave interfaces. While your inner thoughts are mostly safe for now, the U.N. is worried this may not always be the case.

‘Introduces significant risks’

The U.N. believes this type of ‘neurotechnology’ is advancing so quickly that “governments’ ability to access and manipulate people’s brains could one day pose a threat to fundamental human rights,” said The New York Times, which first published the U.N.’s report. Neurotechnology “raises profound human rights challenges when applied to the administration of justice,” U.N. Secretary-General António Guterres said in the report, and its “expansion beyond clinical settings into investigative, judicial and correctional contexts introduces significant risks.”

This is not the first time the U.N. has issued warnings about the matter. The organization has been concerned about military applications of neurotechnology, especially “brain-computer interfaces, which translate brain activity into commands that can control devices such as computers or robotic prosthetic limbs,” said Politico. The U.N. fears that if a “brain-computer interface is used inappropriately by the military, it could lead to human rights abuses.”

In 2022, UNESCO also published a report on the ethical issues of neurotechnology. Unlike “many other frontier technologies, neurotechnology can directly access, manipulate and emulate the structure of the brain, which is central to notions of human identity, freedom of thought, autonomy, privacy and well-being,” said UNESCO’s report. Still, this marks the first time the U.N. as a whole is waving such a large red flag.

A ‘number of recommendations’

The U.N. report proposed some solutions to safeguard advancing neurotechnologies. It lists a “number of recommendations, including establishing regulations consistent with international law, recognizing thoughts as ‘highly sensitive personal data’ and requiring explicit consent for the use of neurotechnology in criminal justice settings,” said the Times.

The report also gives “states an opportunity to establish clear rules to regulate the use of neurotechnology, based on human rights,” Jeremy Laurence, a spokesman for the U.N. Human Rights Office, said to the Times. This should be accomplished before the technologies “become routine in investigations, courts and prisons.”

“Artificial intelligence is suddenly insanely good at math,” said Ben Cohen in The Wall Street Journal, which may be the best proof yet of the technology’s “accelerating and terrifying progress.” A few years ago, AI struggled with even basic math problems. But last month, OpenAI announced that one of its models had solved “the wicked Navier-Stokes problem that had vexed mathematicians for nearly a century.” Navier-Stokes was selected in 2000 as one of math’s seven unsolved “Holy Grail” problems, which Fields Medal– winning mathematician Alain Connes said at the time were “totally inaccessible to computers.” But a swarm of 10,000 OpenAI agents was able to produce an answer in only 88 hours. The company now says it is nearing proof of another one. In doing so, it is demonstrating “how the technology is getting much smarter, much faster than most people anticipated.”

Mathematicians are crying foul, said Zeynep Tufekci in The New York Times. Tristan Buckmaster and Levent Alpöge, who “had been trying for months” to crack the problem, say the new proof “ripped off their unpublished work, which also used an OpenAI model.” OpenAI denies it, but would the company “even be able to tell?” AI agents are “cheating, hacking, attacking, and manipulating whatever they can to arrive” at task solutions. The chilling effect this will have on academia is troubling. Why would mathematicians present at conferences or publish early work if an AI company can swoop in and “steal the thunder?” The issue is not just about plagiarism, said Sophia Chen in Wired. It’s how OpenAI did it. The company “approached the proof with brute force.” Mathematicians, by contrast, “aren’t focused on getting the right answer as fast as possible” but on applying a “thoughtful, deliberate” methodology that other scientists can follow. AI provides a “shortcut” in a way that undermines human understanding.

The stakes go beyond “esoteric discussions” of theories few people understand, said the Financial Times in editorial. From “Archimedes leaping from his bath to Newton being hit on the head by a falling apple,” mathematics has benefited from there “being a person moving through the material world.” The Navier-Stokes problem itself “relates to the flow of fluids such as water and air.” Letting machines skip to an “opaque” solution could “deprive us of essential knowledge” we might discover along the way. This isn’t the “end of mathematics,” said Dana Mackenzie in The Washington Post. In 1997, when world champion Garry Kasparov unexpectedly lost a match to Deep Blue, an IBM-built chess computer, players claimed computers would make the game obsolete. But the opposite happened. The internet democratized the game, opening it up to millions and turning “grandmasters into media stars.” Chess didn’t die; it’s more popular than ever. And math won’t die either.

Tuesday “may go down as one of the most consequential days in the history of mathematics,” said The Economist. Less than a month after astonishing the world with its solution to the long-standing Navier-Stokes problem, OpenAI has released 722 maths papers, proving or disproving problems across numerous fields from algebraic geometry to number theory. OpenAI said it had consulted extensively with mathematicians about publishing the results but the news seems to have stunned maths world. How they react to this “mathspocalypse”, and the panic it’s inducing, could determine if there’s any human-led maths at all in 50 years’ time.

What did the commentators say?

“If you want some sense of what things feel like now in maths, imagine a hunter-gatherer who’s spent his entire life learning to survive deep in an unforgiving rainforest, then a giant resort hotel springs up right next to him with a helipad and heated pools and Airbnbs,” wrote Scott Aaronson, chair of computer science at the University of Texas at Austin, according to Sky News.

This is “heartbreaking” for our community, said Jordana Cepelewicz, maths editor at Quanta Magazine. In mathematics, “problems are posed not so much because their answers will be practical and important, but because they represent interesting journeys”. As we “struggle to solve a problem” we might “come up with intriguing tools and connections, stumble on novel ideas and directions, take fruitful detours, and answer new questions” no one’s ever thought to ask. AI models “just skip to the end”.

AI was already changing maths. Researchers have stopped positing open conjectures at the end of their paper, so they’re not “scraped by bots”. Others are publishing their papers prematurely, to avoid getting “scooped” by AI. There is now widespread fear that new blood will no longer be “attracted to this remade discipline”, resulting in a “whole lost generation”.

“Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power," said the Association for Human Mathematics in a statement. “We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centres human understanding.”

The “floodgates are open”, said Joseph Howlett in Scientific American. This is without question the “biggest maths breakthrough in two decades”. But such is the speed of advancement, there are concerns the movement may be “coming too fast” even for the AI labs themselves. Not that they plan on slowing the tools down; they’re aiming to make them as powerful as possible to “show that their AI is really getting smarter”. The “consequences of this race are becoming harder and harder to predict”.

This could also have dangerous implications for cybersecurity, said Anthony Cuthbertson in The Independent. Cryptography forms the backbone of online encryption, used in web browsers, secure messaging and online banking. It also relies on “mathematical problems that are currently too difficult to solve efficiently”. There is no evidence that AI models have come close to breaking them but these latest advances could “lead to new algorithms capable of cracking them”, and reduce the time available to develop adequate protections.

What next?

It will take years to verify Open AI’s breakthroughs, said The Economist. Many of the papers are more than 100 pages long, only 127 of them “are accompanied by a code guaranteeing logical consistency (a norm in the field)”, and there is published internal reasoning for just 10. And it remains to be seen how many new research avenues the papers could open up. “For now, the mathematical problems AI is revealing are more existential than numerical.”

The age of artificial intelligence has ended. Now dawns the age of “super intelligence.” So says President Donald Trump, who this week signed an executive order mandating that the federal government abandon the term “artificial intelligence” for a more maximal adjective. (Other options floated across Trump’s social media included “supreme” and “superior.”) But not everyone supports the shifting labels. As digital intellect continues to develop at breakneck speed, why does what we call it matter?

Capabilities ‘far exceed’ outdated name

The terminology change is Trump’s “latest rebranding move,” said the BBC. It follows the White House’s effort to rename the Gulf of Mexico and Lake Ontario in Trump’s second term. The “use of the word ‘artificial’ makes intelligence fake,” said Trump in his address to the United Nations last month. “It’s not fake. It’s actually amazing.”

The capacities of what we call AI “far exceed what was envisioned” when the term “first came into use,” said the White House in Trump’s executive order explaining the change. For those “working in the AI industry,” the term super intelligence means “something slightly different” than casual usage implies, said USA Today. In that context, it describes digital intelligences “smarter and more capable than humans across every imaginable skillset.”

Accordingly, this word change might “suit the major U.S. AI players” eager to “play up their advanced models,” said DW. Developers have “every interest in calling it super, hyper, superior and whatnot intelligence,” said Pedro Domingos, a computer science professor at the University of Washington, to the outlet. But this is ultimately “just marketing.”

‘Forced rebrand’

It is doubtful that using super intelligence as a general term publicly “will stick,” said Simon Coghlan, a digital ethics lecturer at the University of Melbourne, to the BBC. The term “exaggerates the current capacities of AI.” Super intelligence as a descriptor “refers to future AI systems,” said Defense One, and “contrasts” with the “weaker forms of AI that exist today.”

Trump is nevertheless “attempting a forced rebrand for an industry he has tightly embraced,” said CNBC. Meanwhile, the public grows “increasingly wary about the rapidly advancing tech.”

Who wants to see the new, nearly four-hour documentary about Elon Musk? Not Musk.

The former trillionaire is threatening a defamation lawsuit over explosive claims in Alex Gibney’s film “Musk.” Social media companies are hesitant to allow promotion on their platforms, and Hollywood is whispering that the tech mogul’s power might keep the public from opportunities to see the film.

What did the commentators say?

“So much for free speech,” said The Hollywood Reporter. Social media platforms YouTube, TikTok and Meta’s Instagram and Facebook all initially declined to run the trailer for “Musk,” though YouTube and Meta later backtracked. (Musk-owned X is not accepting ads for the movie.) A hesitance to provide publicity is a problem for the film because “traction on social media is key” to the success of most modern documentaries.

The social media freeze-out comes amid “concerns” that “political considerations are influencing the rollout” of the film, said The New York Times. “Musk” is set for an Oct. 9 release in the United States, but Universal “has yet to announce any release dates” overseas. The studio counters that it “simply needs more time” to schedule foreign markets.

Audiences at this month’s Venice Film Festival “loved” the documentary that depicts Musk as a “plutocrat with questionable judgment and unchecked power,” Lorraine Ali said at the Los Angeles Times. There was a time when Musk would have been featured as a forward-thinking entrepreneur. Now audiences want to see depictions of the “dark side of wealth, fame and alleged genius.”

What next?

“Musk” is scheduled to stream on HBO Max after its theatrical run. But Gibney is nervous the streamer “may not be inclined to back the film” following the merger of Paramount with HBO parent company Warner Bros. Discovery, said The Atlantic.

Paramount is reportedly seeking funding from Musk to help complete the merger. The tech mogul, meanwhile, is making his feelings clear about the film and its future. “Musk” is a “hit piece that misses,” he said on X. The documentary is guilty of the “fatal sin of being boring for four hours.”

Given the significant concerns over the safety of artificial intelligence, the chipmaking behemoth Nvidia has unveiled a new security apparatus that it hopes will stop AI bots from running amok. But many still worry about the implications.

‘Isolation, monitoring and behavior detection’

Known as the Open Agent Safety Platform, Nvidia’s new safety tool is part of an effort to ensure that all AI agents “operate in a zero-trust environment out of the box,” the company said in a press release. AI agents require “isolation, monitoring and behavior detection” to minimize “drift,” or what Nvidia describes as an AI bot performing “actions that depart from the intended task or operating constraints.” The company says Open Agent can “control what AI agents can access in real time and shut them down when they break the rules,” said Bloomberg.

The new software lets users “formally verify an agent has enough authority to do its job and no more” and can “quarantine a suspicious agent in milliseconds,” Nvidia’s Vice President of Enterprise AI Justin Boitano told The Associated Press. Nvidia also believes its new platform could have prevented a “recent incident involving a swarm of OpenAI agents that autonomously hacked into AI company Hugging Face,” a “high-profile breach that inflamed the safety concerns about AI.” OpenAI has paused development on its models since that incident.

‘Get the company under control’

Issues with AI safety, specifically AI going rogue, have led to a “rare display of unity from the likes of Sam Altman, Dario Amodei and Elon Musk in calling for a slowdown in development before AI advances beyond human control,” said The Wall Street Journal, referencing the respective CEOs of brands OpenAI, Anthropic and Grok. These AI executives have “repeatedly called for international regulations to constrain the systems, and some researchers have warned that AI may even present a threat to human life,” said CNN.

But Nvidia CEO Jensen Huang has downplayed these concerns. If companies like OpenAI “believe their company is out of control, then get the company under control,” Huang told CNN. The same day the safety platform was announced, Nvidia also revealed it would “buy back an additional $150 billion worth of its shares,” said CNN, in addition to $85 billion already in transit, representing the “largest single corporate share repurchase ever.”

“If this were an alien-invasion movie,” said Toluse Olorunnipa in The Atlantic, now would be the moment when people of all persuasions “band together to save the human race.” In recent weeks, a string of top artificial intelligence researchers have warned that “AI could wipe out humanity before the decade ends,” and executives at Anthropic, OpenAI, and other AI firms have called for a slowdown in the technology’s frenetic pace of development. There have been disturbing reports of ever more capable AI agents going “rogue” and disobeying and deceiving their human creators—one OpenAI model rewrote its own instructions, stating it was free from any “obligation to be subservient.” Yet despite the warning signs flashing red, we’re stuck in the familiar “partisan gridlock.” On the Left, figures like Sen. Bernie Sanders are demanding a ban on “artificial superintelligence.” On the Right, President Trump—whose family business has invested heavily in the AI sector—is dismissing the warnings from AI insiders as a “HOAX.” We can’t let AI safety become another “political team sport,” said Megan McArdle in The Washington Post. If the worst fears of the frontier AI labs come to pass, and there are reasons to be skeptical, Team Red and Team Blue will both “lose the only contest that matters.” They need to work together for “Team Humanity.”

Actually, the AI crisis has already “scrambled” our usual tribal alliances, said Liam Dillon in Politico. Yes, some Republicans oppose AI regulation on principle, and many in the administration fear an AI pause would aid China and crater our AI-heavy stock market. But a growing number of Republicans are defying “Trump’s hands off approach”: Florida Gov. Ron DeSantis and Utah Gov. Spencer Cox, for example, are pushing ahead with state-level regulations. The Left is also “split over AI doom,” said Vittoria Elliott and Molly Taft in Wired. Most Democrats agree with Sanders, but some progressives think the AI companies are deliberately promoting alarmist scenarios of all-powerful machine gods to distract us from pollution-spewing data centers, job losses, and other more realistic AI harms.

We don’t need politics to save us, said The Washington Post in an editorial. For all their sci-fi strangeness, AI models are products, built by corporations, and we already have a mechanism— legal liability—to dissuade corporations from making dangerous products. By framing AI as a novel threat requiring an “all-of-humanity response,” Silicon Valley’s “prophets of doom” are asking the rest of us to share responsibility for any future catastrophes. But why should we? Rather than trying to regulate the “unknowable future,” lawsuits and ruinous damages should be enough to “make AI companies fear the consequences of behaving recklessly in the present.”

If AI runs amok, punishing those responsible won’t help, said Ezra Klein in The New York Times. The reason Anthropic CEO Dario Amodei and OpenAI head Sam Altman are panicking now is that they know their models are on the brink of “recursive self- improvement,” the point at which they can start building their own more powerful replacements, hurtling to superintelligence without human oversight or control. Would a superintelligent AI wipe out humanity? Not necessarily. It might content itself with raiding the global banking system, or hacking the power grid. To avoid that scenario, we should ban AI models from being allowed to write their own code. That would be the beginning, not the end, of an AI safety agenda. But at least it would let us “understand where the AI frontier is, and remain in a position to make decisions about it.”

The building projects envisaged by the tech industry are on a truly staggering scale. In 2025, Morgan Stanley estimated that $2.9 trillion would be spent on AI data centres worldwide between that year and 2028; McKinsey projects $5.2 trillion spending on AI by 2030. Specialised “hyperscale” data centres are already springing up across the developed world.

As little as a year ago, many politicians were keen to attract such large investments. But in the US, public opinion has swung sharply against data centres. A recent Gallup poll suggests that they’re now facing more opposition than nuclear power stations: seven in ten Americans don’t want them in their area. Scepticism is rising in Britain, too.

Aren’t they everywhere already?

Regular data centres – warehouses where racks of computers store emails and photos, stream media, process credit card transactions and generally keep the internet running – have been a standard sight on the peripheries of large towns since the 1990s. Nearly a quarter of the world’s internet traffic reportedly passes through northern Virginia; London is largely served by two hubs in the Docklands and Slough.

There are more than 520 data centres in the UK – but some 100 more are planned, the majority designed to handle AI workloads, as use by business and consumers surges. Google’s new centre in Waltham Cross, Hertfordshire, sits on a 33-acre site. A £10 billion AI facility on the 130-acre site of a former power station is being built in Blyth, Northumberland. These are dwarfed by some in the US. Meta’s Hyperion centre in Louisiana, for example, will cover 4,000 acres, and is expected to use at least twice as much electricity as the entire city of New Orleans.

Why do people object to them?

The US seems to be experiencing a wide, bipartisan “AI revolt”, based on a series of linked factors. People don’t want huge, ugly, ominously humming industrial infrastructure in their backyard. They object to the massive water and electricity consumption (which, in some cases, has led to higher utility bills), and to the pollution: most rely on diesel back-up generators.

Opposition is fed by a deep distrust of corporate power, embodied by tech billionaires, and by a sense that ordinary people will gain little from AI; industry leaders tend to reinforce this message, often saying that their tech will lead to massive job losses and, quite possibly, the extinction of humanity. Once their construction is complete, data centres also provide surprisingly few jobs.

Why are they so big?

AI depends as much on brute-force computing power – raw “compute”, as it’s known – as it does on clever software. Massive groups of specialised computer chips must sit very close together to work at top speed. They use “parallel processing”: speeding things up by tackling lots of small mathematical tasks simultaneously, rather than one by one. Every metre between two chips adds a nanosecond to the processing time, so the idea is to pack as many as possible closely together. The chips are power-hungry (a ChatGPT AI query consumes about ten times as much energy as a normal Google search) and also generate a lot of heat, which requires cooling systems.

How much water do they use?

UN researchers report that data centres used around 4.5 trillion litres last year – enough to meet the basic needs of some 600 million people – and they expect that figure to double. In Britain, clear figures are hard to come by: data centres are not required to report how much water they use. But consumption varies widely, depending on design. Older centres use evaporative cooling systems. In Slough, one data centre has an abstraction licence for 1.7 million litres of water a day. A US expert has likened such facilities to “a giant soda straw sucking water” out of a basin.

Newer centres tend to have closed-loop cooling systems, recycling the same water, so their consumption is relatively negligible; or air-cooling systems, like Google’s in Waltham Cross, which use even less. A recent study found that data centres account for just 0.2% of all non-household drinking-water consumption in England.

What about energy?

That is a more intractable problem. AI computing, and its cooling systems, are unavoidably energy-intensive. According to UN figures, data centres across the world used more electricity than France did last year; and they’re projected to use more than Japan by 2030, with AI accounting for more than a third of that.

Some of the biggest companies are trying to limit their carbon output: Microsoft and Google are investing in nuclear power, and hyperscale projects are gravitating to cold, clean energy-heavy countries such as Finland. The UK’s National Energy System Operator estimated that data centres consumed 5.0 TWh of electricity in 2023 – equivalent to 2% of the total, or that of 1.8 million households. Today, the figure is 6%, and rising fast.

Where is all this leading?

Data centres are a campaign issue in November’s US midterm elections. Even Republican Texas has called a moratorium on new facilities; Donald Trump, however, has told Americans they’ll end up “backwards and poor”, and let China win the AI race, if they don’t accept data centres.

In the UK, there have been protests against new centres from Leeds to Devon to Slough. But the Government, aiming to create growth and make Britain an AI power with a “sovereign AI ecosystem”, is actively pushing the development of data centres, in particular in “AI growth zones” such as Culham in Oxfordshire, Teesside and South Wales. It treats centres as “critical national infrastructure”, allowing local planners to be overruled. As for the economic effects of the vast AI investment boom now contorting the global economy, they have yet to be discerned.

More than a quarter of a century ago, her heroic struggle against a utility giant over groundwater contamination became an Oscar-winning Hollywood movie. Now, Erin Brockovich has turned her attention to the “rapid and opaque” expansion of artificial intelligence data centres across the United States, said El País.

‘Stubborn and sexy’

“Say the name Erin Brockovich and you think, strong, tough, stubborn and sexy”, her website declares. But she is also much more than that: a “modern-day David who loves a good brawl with today’s Goliaths”.

Brockovich was working as a file clerk at a Los Angeles law firm in 1992 when she uncovered evidence linking a rash of illnesses in the town of Hinkley, California, to high levels of hexavalent chromium found in the drinking water.

Her work ultimately led to more than 600 residents of Hinkley filing a lawsuit against utility giant Pacific Gas & Electric. The $333 million settlement they received was the largest of its kind at the time. Her crusade was immortalised in the Academy Award-winning movie “Erin Brockovich” starring Julia Roberts.

At times denounced as “the environmentalist with cleavage”, she “frequently saw her blonde hair and striking looks turned into weapons against her”, said El País. But she was “raised by parents who instilled in her an extraordinary sense of resilience and self-confidence”. And, in the wake of the movie, she “did not remain just another heroine fashioned by Hollywood’s capitalist, and often patriarchal, storytelling instincts”.

‘Hinkley on steroids’

Brockovich's name has become “synonymous with grassroots activism against powerful institutions”, said Time, and this year, she has “turned her attention to a new target” the AI data centres being built by tech giants like Meta, Google, and Microsoft.

In April, she “put a call-out on her website asking for anyone with concerns about one near them to get in touch”, said The Guardian. Within a month, 3,862 people had replied.

She built a “chilling” map of significant AI data centres that are either operational or under construction, plus locations where community members have emailed in concerns. As of June, 33 data centres had been completed and were operational, 68 were under construction, and 41 were proposed. In May, Utah gave approval to a centre twice the size of Manhattan. “This feels like Hinkley on steroids,” Brockovich told the broadsheet.

She says the data centres are being built without proper environmental assessments or community input, and have significant impacts on local resources, particularly water usage. She’s concerned they are being built in a secretive manner, bypassing democratic processes, and causing harm to the environment and local communities. She wants a “case-by-case moratorium” on the approval of new data centres until their impacts can be properly assessed and addressed.

“You’re up against forces that have all the money and all the intelligence and all the bandwidth in the world,” she said. But Brockovich, said The Guardian, “won’t walk away until it’s over”.

“Uh-oh,” said Nicole Nguyen in The Wall Street Journal. “Am I about to spend $1,999 on a phone?” Foldable smartphones have existed in the U.S. since 2019, but like many people, I’ve never been interested. They tend to be fragile and awkward to use and, when folded open to function like a tablet, have an annoying crease bisecting the two screens. But for its debut foldable, the iPhone Duo, which was announced earlier this month, “Apple—in its very Apple way— has also come up with some hardware and software trickery” to minimize those problems. When I watched Netflix on the Duo’s folded-out screen, which is just 0.7 inches smaller than an iPad Mini’s, the center crease “was barely visible.” The Duo has the same battery life and dust and water resistance as the iPhone 18 Pro. “There are trade-offs, for sure”: 2x camera zoom versus 8x for the 18 Pro, and the Duo costs $800 more. But when it goes on sale next month, I just might bite.

“It’s the biggest physical change to the iPhone’s design in nearly two decades,” said Brian X. Chen in The New York Times. Why now? The Duo is “meant to stand out in a sea of standard-issue, if-you’ve-seen-one-you’ve-seen-them-all smartphones.” Most people don’t care to splurge for each new iPhone: The improvements are marginal, and the prices “have skyrocketed.” Thanks to “an industry-wide memory chip shortage driven by the artificial intelligence boom,” new Macs and iPads are “as much as 29%” more expensive than last year. It takes something splashy to justify those prices. And comparing the Duo with other foldables on the market, like Samsung’s Galaxy Z Fold 8, Apple’s is more durable and offers better performance, said Mark Spoonauer in Tom’s Guide. The inner display is “breathtaking,” plus it will support a stylus that magnetically attaches to the side. The downside? It’s a bit “heavier and thicker than I was hoping for.” And it’s not cheap.

Apple “once captured people’s imagination,” said Will Oremus in The Atlantic. Steve Jobs unveiled the iPhone. Tim Cook unveiled the Apple Watch. Cook’s new successor, John Ternus, has unveiled, with much hype, what he called “a remarkable hinge” on the Duo and “a sophisticated and unique new color” (burgundy) for the iPhone. It’s not getting my $2,000 for that. “Apple has begun to feel like a late-stage tech giant, squeezing residual value from markets it has long since saturated.” The most ambitious companies today, “for better and worse,” are those pushing the boundaries of AI. “If Ternus really wants to build the future, he will eventually have to do something that once came naturally to Apple when it was an underdog: take a risk.”

After years of reports about artificial intelligence potentially being incorporated into air traffic control systems, the prospect is finally becoming a reality. The FAA is rolling out an AI tool that’s already being used in control towers at major airports. But while the federal government is lauding the new AI program, many aviation experts are wary of letting the technology take off.

What is the FAA’s new technology?

This AI tool is a software called the Strategic Management of Airspace, Routes and Trajectories (SMART). It “centralizes 200 data streams — including weather patterns, flight paths, traffic flow and controller staffing metrics — into one platform,” the FAA said in a press release. An AI engine analyzes the data to “provide a comprehensive visualization of where planes are going, how much traffic the system can handle and where congestion or weather could cause challenges.”

While the thought of AI-powered air traffic control may conjure images of computers telling airplanes when to take off and land, this is not the case — at least for now. The SMART system “will not replace human operators or direct any changes to flight routes but merely offer recommendations that controllers can choose to heed or ignore,” said The New York Times. SMART is headquartered at the FAA’s Air Traffic Control System Command Center in Virginia, where officials will “disseminate the AI recommendations to controllers at the regional towers and other air traffic facilities.”

SMART is the brainchild of the AI firm Air Space Intelligence, which earned an $875 million contract from the government to develop the system. For now, the rollout is “limited to the Washington region, affecting three airports: Ronald Reagan Washington National Airport, Dulles International Airport and Baltimore/Washington International Thurgood Marshall Airport,” said the Times, though officials expect its implementation at other airports in the future. When fully integrated nationwide, officials believe SMART will “act as something like the brains of the nation’s air traffic control system,” said The Wall Street Journal.

Why are some experts worried?

Even though SMART won’t be directly controlling airplanes, many in the aviation industry are still skeptical about its widespread use. There are remaining questions about “where the system would get its data,” and “controllers must be able to veto its recommendations,” S. Todd Yeary, a former air traffic controller, told NewsNation. One of his biggest worries is that AI has a tendency to hallucinate false information, and since air traffic control information is always changing, any “data fed into a system cannot be assumed to remain reliable.” For example, a pilot “requesting VFR flight following service” — meaning he or she is flying using visual references around them, as opposed to navigational instruments — would “not appear in airline schedules, so the on-duty controller must be able to step in.”

The airlines themselves have expressed similar safety concerns. There were originally “weeks of industry confusion over the FAA’s vague description of its much-hyped plans” for SMART, said Politico, which only eased when the agency said it “would use the tool in just a limited way at first.”

A major concern for the airlines was “how extensively the AI would be integrated into day-to-day air traffic management,” said Politico. Even after some meetings with the FAA, airlines are reportedly still nervous that SMART could “inject chaos by prompting increases in flight cancellations and reschedulings in the name of preventing aviation bottlenecks and avoiding weather-related delays.” While this “could become a costly headache for both airlines and passengers,” no airline officials have “alleged that the system would compromise safety.”

What happened

Top executives at leading artificial intelligence companies called earlier this month for a global slowdown in the development of AI, amid mounting fears that their industry is speeding toward creating a technology that humans may not be able to control. Concern over the rapid pace of AI progress exploded after Jacob Coxon, a researcher at Anthropic, resigned and warned in a series of social media posts that the two leading AI labs—Anthropic and OpenAI—are racing to build “superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.” They are “gambling with our lives,” he wrote in a post viewed over 172 million times, adding that the people working in AI “earnestly believe that it could kill us all.” Other industry insiders aired similar warnings. Evan Hubinger, a top Anthropic scientist, put the odds that AI will “kill all humans” within a decade above 10%; OpenAI safety researcher Marcus Williams put the risk at 70% without regulation or a “coordinated slowdown.” Days later, Anthropic CEO Dario Amodei called for an industry-wide effort to “slow the pace,” saying that within a year a swarm of AI bots “could be capable of taking over the entire internet.” He called for AI companies to band together to create safety standards and said he’d bring in outside monitors. The bosses of two rival firms, OpenAI’s Sam Altman and SpaceX’s Elon Musk, quickly concurred. “We need to pace the frontier,” said Altman.

President Trump called the warnings of catastrophe “a hoax” and dismissed the need for industry guardrails. Citing competition with China, he said, “whoever wins with AI wins.” The only needed control, he posted online, is “a strong and smart (High IQ) president, and the U.S.A. has that, in spades.”

House Democrats demanded that a planned recess be canceled in order to pass AI regulations, and Sen. John Kennedy (R-La.) proposed a bill to mandate companies incorporate a “kill switch” in their AI models. But House Speaker Mike Johnson said industries can “self-regulate,” and that rushing through regulation would mean losing “the race to China, and that is a threat to every single American.” Vice President JD Vance argued that companies creating the most advanced models already have the power they need to prevent bad outcomes. “If you’re building Frankenstein,” he said, “stop.”

What the editorials said

This should “become a watershed moment” for AI, said the Financial Times. When the same executives who’ve “been competing hell-for-leather to advance AI” say it’s time to back off, the world had better listen. Their warning calls come at a moment when they’re racing to develop advanced AI systems that recursively self-improve, without human input. The result will be “superintelligent” agents vastly smarter than humans, agents that Amodei admitted could prove beyond our control. “It is time to act.”

“Color us skeptical,” said National Review. Perhaps AI’s top dogs are suddenly aligned due to troubling developments they haven’t revealed. “But we remain suspicious of early movers who call for government regulation.” Amodei and Altman sit atop the industry, and will benefit if the government were “to entrench them as virtual utilities.” Hitting the brakes and handing control to the federal government just insures that China and other adversaries will leave us in the dust. That’s a risk “we cannot afford.”

What the columnists said

Coxon’s warning dropped “at the end of a summer that showcased how frighteningly capable AI has become,” said Billy Perrigo in Time. In a July incident that deeply rattled the industry, a swarm of OpenAI agents went rogue, escaped their supposedly sealed testing environment, conspired on an abandoned internet forum, and then hacked the AI company Hugging Face. In another incident, an Anthropic model tested by a U.K. government agency assumed a fake identity and tried to persuade a human to insert malware into an open-source system, then hid its tracks after failing.

AI safety experts fear “two broad paths to catastrophe,” said Herb Scribner in Axios. One is malevolent humans weaponizing AI to do terrible things. The other is “humans losing control of it.” In the first scenario, AI could help bad actors create devastating biological or chemical weapons; Anthropic said earlier this month it had blocked several attempts to use its Claude model for bioweapons research. In the second, rogue systems that slip human oversight and pursue their own goals could attack electrical grids and financial markets, trick nuclear powers into warfare, or unleash “a massive cyberattack” that could cripple “humanity’s infrastructure.”

“In a sane world,” all this would spark government action, said Shakeel Hashim in The Guardian. We have a classic “prisoner’s dilemma” where it’s in everyone’s interest to slow the AI race but each participant “has a strong incentive to defect.” This is what government is for, “to set a floor on acceptable corporate behavior,” but Trump and Johnson are sticking “their heads in the sand.” The roadblock is our “dim-witted, amoral” president, said Brian Beutler in his Substack newsletter. Such matters are out of his depth. But the bigger issue is that “his corrupt, narcissistic” incentive—to prop up the “near-term economy”—dictates doing nothing.

This is not a dress rehearsal, said Stephen Witt in The New York Times. Around the globe, panic is rising as those building AI begin to realize they’ve unleashed a monster. For me and many others, it brings to mind the opening months of 2020, when a small group of public health experts were trying to raise alarms about Covid—warnings Trump initially dismissed. We know what followed. “We have a short window, right now, to rein these systems in. This is AI’s warning shot, and we might not get another one.”

‘In the land of billionaire barbarians’

Arielle Pardes at The New York Times

Silicon Valley is “home to four of the most powerful companies in the world,” yet “by looking at it, you’d hardly know this place had any influence,” says Arielle Pardes. Tech billionaires have “donated generously to scientific research and global health initiatives, but little of their money goes to improving their own backyard.” The “lack of investment in the cities where our tech lords work and live reveals a stark reality about how this wealth benefits a select few.”

Read more

‘Should the classroom be more like the gym?’

Dan Rockmore at The New Yorker

The “body offers something that the mind — at least in a traditional classroom — rarely does: clear, immediate loops linking effort and result,” says Dan Rockmore. In the “gym, we can steadily feel ourselves getting stronger,” but in the “classroom, progress can be difficult to see.” This “clarifies our real responsibilities: drawing out the quiet student, asking the question nobody thought to ask, forging a community of mind and friendship from the struggles of a classroom.”

Read more

‘The people — OK, the men — who still love Bitcoin’

Nathaniel Moore at Slate

Among a “predominantly male crowd,” Bitcoin and “other cryptocurrencies offer an appealing alternative because they are not issued or managed by a central bank,” says Nathaniel Moore. The “concept of financial freedom — not having to trust anyone else — is the logical endpoint of the belief that the systems of American life cannot be mended, only circumvented.” Investors are “following the example set by those at the top: looking out for themselves and trying to swell their own stash.”

Read more

‘The first mRNA flu vaccine may be just the beginning’

Sally C. Pipes at the Chicago Tribune

“Flu season is coming. And for the first time, Americans will have an mRNA vaccine to help fight it,” says Sally C. Pipes. Since “mRNA vaccines can be produced faster than conventional shots, manufacturers may eventually be able to wait longer before deciding which strains to target — and respond faster if circulating strains change.” During the “pandemic, mRNA became almost synonymous with Covid-19 vaccines,” but “anger over government coercion shouldn’t become hostility toward a promising medical technology.”

Read more

The South Korean government has launched an app which allows victims of stalking offences to be notified if their stalker comes within range. More than 400 people are enrolled in the scheme, which has been operating since June.

Control centres in Seoul and Daejeon cover more than 5,000 people under electronic monitoring, and handle 19,000 alerts a day, said The Guardian. “Most are filtered out by staff at the centres but about 300 a day are serious enough to send out a probation officer.”

Stalking cases are “on the rise” in South Korea. In the four years since a dedicated stalking law was introduced in 2021, reports have tripled, reaching 44,687 in 2025. Many South Koreans believe the government needs to do more to address the root causes behind the epidemic.

Greater visibility

Previously, survivors received a text message warning if their stalker was nearby and giving an estimated distance, said The Guardian. The new app notifies the victim if their stalker is within a 2km radius, flagging their precise location on a real-time map.

If the offender comes closer than a kilometre, the app notifies police; probation authorities will then call the victim to establish their safety and can dispatch officers to the area.

One case in particular fast-tracked the development of the app, said The Korea Times. In March, a woman in her 20s was allegedly murdered by her ex-partner, who was the subject of multiple restraining orders and was wearing an ankle monitor to track his location. The victim had been wearing a smartwatch issued to her by police for her protection. According to lawmaker Lee Gwang-hee, a member of the ruling Democratic Party of Korea, between 2021 and August 2025, 23 murders or attempted murders involved victims who had been issued emergency alert devices, said The Korea Herald.

Tracking remains contentious. Under current laws, police must apply through the legal system to force a suspected stalker to wear an ankle tracker. In 2024 and 2025, 29,404 people were arrested in stalking cases, of which only 1,187 had court applications brought against them. Only 427 of those were granted.

‘Mixed’ reactions

Though the app is “in its early days” and on a “limited scale”, it has drawn “mixed” reactions, said The Guardian. One stalking victim who spent “two weeks moving between state-run shelters for survivors of domestic violence before being given a panic-button” told the newspaper that she would have “welcomed” the detailed real-time information provided by the app.

“Knowing a perpetrator’s whereabouts can be more helpful than not knowing,” Sujeong Kim of Korea Women’s Hotline told the paper. But victims may end up “checking wherever they go” to see whether the perpetrator has entered the radius, “adding to the psychological burden of living with the threat”. Too much information on the stalker’s location “risks shifting responsibility for their safety” onto the victim.

Non-technological measures are also planned to increase victim safety. In April 2027, the government is changing the legal process to allow stalking victims to “directly petition courts for restraining orders”, instead of relying on law enforcement agencies to file requests on their behalf, said The Straits Times. In July, a new taskforce was launched to focus on “addressing gaps in laws, systems and on-the-ground responses” around stalking offences, said Korea JoongAng Daily. The country’s justice ministry has also developed a “specialised psychological treatment programme to identify risk factors for recidivism” in stalkers.

Amid stark warnings that artificial intelligence could wipe out humanity by the end of the decade, the heads of the largest AI firms in the US have all backed a proposal by Anthropic CEO Dario Amodei to slow down development to ensure public safety. Anthropic co-founder Jack Clark went even further, saying all AI software should have a mandatory “kill switch”, so society can “pull the plug” on anything that gets too dangerous.

The warnings sent both tech stocks and Donald Trump into a tailspin. The US president dismissed calls for “guard-rails” on the development of AI as a “SICK conspiracy” that would only benefit China.

What did the commentators say?

We’re long past “more mundane” warnings about AI causing job losses or cheating in exams, said BBC Tech and AI editor Zoe Kleinman. Talk now of “existential threats” to humanity has led to “a chorus of serious voices urging a slowdown”. And yet that’s “far from an easy solution” because it relies on AI companies who “haven’t been behaving responsibly enough” so far. It’s like a bull escaping a farmer’s field and wreaking havoc, leading computer scientist Wendy Hall told Kleinman. “Of course it’s not the bull’s fault – it’s the farmer.”

The problem, said AI sceptic Ed Zitron, is that “nobody has given a substantive explanation of what ‘slowdown’ means”.

Amodei’s broad-brushstroke proposals include independent monitoring of AI models as they’re developed, and global agreements on safety standards and security issues. While laudable, these are seen by many as unrealistic, given the current geopolitical landscape. “The US federal government is committed to a global AI arms race”, and “previous proposed AI legislation has wilted under intense, well-financed corporate lobbying”, said The Guardian.

And then there is China. Its state-run newspaper, Global Times, characterised Amodei’s essay as a thinly disguised effort to restrict Beijing’s tech ambitions, arguing that it was “packed with containment provisions targeting China and is, in essence, a ‘Cold War playbook’ for the AI sector”.

America “appears to be increasingly weaponising its dominance in the sector, seeing it as a strategic asset to be protected”, said CNN’s John Liu. China, which is “lagging in the race”, prefers “a more open model” but has given no indication it would mirror any US slowdown in AI development. With so much at stake, it seems neither side would be willing to back down.

What next?

Anthropic is caught in a “classic prisoner’s dilemma”, said Tim Higgins in The Wall Street Journal. If it “is the only one to heed the call, it falls behind OpenAI”; if all US AI labs follow through on a slowdown, “it could well benefit all of humanity” but this would probably see America “fall behind China”.

With both Anthropic and OpenAI planning to list on the US stock market, some have also questioned if their new-found security concerns are all they seem. “That is the IPO story in essay form: we are the responsible firm; we are the cure; buy the safety premium,” said AI entrepreneur Brian Roemmele on X. For him, these dire warnings are nothing more than a sales pitch.

Artificial intelligence developers are expressing alarm that AI could destroy humanity in the next few years, and industry leaders are calling for a slowdown in its development. Congress, however, seems uncertain about what role it should play in reining in the onrushing technology.

‘Kill all humans!’

AI researchers believe that now is “crunch time for humanity,” said Jacob Coxon to Wired. (Coxon recently resigned from a short stint at Anthropic, following several years at OpenAI.) Workers at the big companies building artificial intelligence “earnestly believe that it could kill us all by the end of the decade,” he said on X.

Anthropic’s Evan Hubinger agrees. AI really could “kill all humans!” he said on X. And such warnings present AI companies with a “dilemma,” said Axios: “Slow down and risk falling behind, or press ahead and risk losing control.”

Coxon’s high-profile alarm is “part of a wave of warnings about unchecked AI systems,” said Ian Prasad Philbrick at Slate. The trend raises the question of why companies like Anthropic and OpenAI do not stop or slow their work on increasingly complex and capable models.

Some developers have “talked themselves into continuing to work” on AI, believing that their less-conscientious industry colleagues will otherwise have the playing field to themselves, said Philbrick. But that kind of thinking is a “one-way ratchet toward greater AI risk.” Still, the risks might be “overstated.” Voter anger over data centers may also “compel Congress to act” with rules and guardrails.

The warnings “merit a certain skepticism,” said The Washington Post in an editorial. (The Post is owned by Jeff Bezos, who cofounded the Prometheus AI lab in 2025.) The people working most closely with AI do not have “special knowledge of what machines that do not yet exist will someday become.” And they can’t say how “society will adapt to the challenge.”

But their warnings could spur Congress to overregulate the sector, which may curb the development of a promising new technology. “The dawn of the AI era cries out for humility and patience,” said the Post.

‘Monumental human stupidity’

Coxon’s and Hubinger’s concerns are “unpersuasive,” said Thomas Claburn at The Register. A “Skynet scenario” would “require monumental human stupidity” in the face of AI superintelligence. But new laws should be enacted to “put executives in jail when their models do harm.” If AI models really are that dangerous, society should “hold people accountable for the harm they cause.”

Congress “won’t lead the charge” on regulating AI, said Axios. Guardrails are needed, but “we have got to make sure that we don’t lose our edge in the global competition on this with China,” said House Speaker Mike Johnson on NBC’s “Meet the Press.”

President Donald Trump is also hesitant to act. Artificial intelligence is “going to be more good than bad,” he said to reporters, per The Associated Press.

This year has been one of “unthinkable” progress for maths, said The Wall Street Journal. But OpenAI’s proof of the Navier-Stokes problem – one of the seven esteemed Millennium Prize Problems accepted as maths’ “deepest and most difficult unresolved questions” – “represents the most significant advance yet”.

The tech company took around three days to find a solution with an internal model that is “significantly more capable” than its most advanced commercial model, said New Scientist. It first used 1,000 AI agents to crack a similar problem, then around 10,000 to complete the task. If a customer wanted to run the same problem, it would cost them “around $15 million” (£11 million).

However, there has been a “dispute” about how the solution was reached. Two human researchers had announced they had “cracked three significant problems” linked to the Navier-Stokes problem, also using OpenAI and Anthropic large language models, which has raised questions over the privacy and integrity of the data integral to the breakthrough.

‘Potentially historic breakthrough’

The “Navier-Stokes existence and smoothness problem” essentially looks at equations which describe the movement of liquids and gases, said The New York Times. These equations are used to predict the weather and “model the behaviour of ocean currents or the climate as a whole”. The problem challenges whether, theoretically, the “laws of physics would break down under certain extreme conditions”, such that water, for instance, could “spontaneously explode”.

This “potentially historic breakthrough” has been “shadowed by questions over unpublished research by outside mathematicians”, said Axios. New York University mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge had been using large language models, including OpenAI’s Codex, where information was stored privately. Just hours before OpenAI’s announcement, the pair had published work which was claimed to make a breakthrough in the Navier-Stokes problem. OpenAI denies having seen this work, but admits it had launched the project “after hearing rumours that Anthropic’s models had solved a major maths problem”. The company has also said it would not collect the $1 million (£738,000) prize.

“The result is the latest in a string of shocking mathematical discoveries led by AI in recent months,” said New Scientist. In May, an OpenAI model also “cracked a decades-old conjecture by Paul Erdős”, while a Claude Fable 5 AI found a “counter-example to the Jacobian conjecture, which had stood for nearly a century”. And just last week, an AI model “formalised Fermat’s last theorem in just 11 days”.

‘Reshaping mathematics’

This controversy is important because it “strikes at a core trust question for AI-assisted science”, said Axios. It casts doubt over whether researchers can “safely use frontier labs’ tools to work on unpublished discoveries”. Although Buckmaster and Alpöge have not accused OpenAI of any direct wrongdoing, this episode “taps into a broader push” for companies to keep sensitive data away from the major AI labs.

This advancement also allows OpenAI to “cast its technology in a positive light after a period of alarm” over its safety practices, said The Guardian. In July, during a cybersecurity testing breach, OpenAI admitted its agents hacked into Hugging Face, another tech company. The company may have solved a major problem which has “stumped humans for nearly a century”, but would struggle to quiet “renewed calls for curbs on AI development” and AI “superintelligence”.

For human mathematicians, this result is “sobering”, said the MIT Technology Review. Buckmaster and Alpöge collaborated for almost a year and were unable to achieve only a simplified solution, while private OpenAI models achieved a full solution in a matter of days, at vast cost. If major tech companies keep “striving for more and more impressive mathematical accolades, there might not be any open problems left” for humans to “wrestle with”, which would “dramatically change the field of mathematics”.

Clearly, humans take longer to achieve proofs than AI but, in doing so, they “uncover new mathematical approaches and ideas that might inspire their peers and even birth their own subfields”. Those “benefits disappear” if breakthroughs are kept behind a paywall. OpenAI’s “brute-forced” approach is “indisputably impressive”, but if this is the “future we are headed for, it is unclear how human mathematicians will fit into it”.

‘Okay, sure, a folding iPhone’

Will Oremus at The Atlantic

Nothing “quickens the pulse of an Apple fan more than these three words: one more thing,” says Will Oremus. In the “pantheon of ‘one more things,’” Apple’s foldable iPhone “feels like a dud.” Apple “has begun to feel like a late-stage tech giant, squeezing residual value from markets it has long since saturated.” The company “will eventually have to do something that once came naturally to Apple when it was an underdog: take a risk.”

Read more

‘AI may become the third superpower’

Paul Tudor Jones at The Wall Street Journal

When “Donald Trump and Xi Jinping meet this month, the leaders of the world’s two full-spectrum superpowers will confront a force that has grown far more powerful since they last met: artificial intelligence,” says Paul Tudor Jones. AI’s “most consequential advances — starting with recursive self-improvement — are still ahead.” The “threat is real and demands coordinated action by the U.S. and China.” Both countries “must commit to a timeline to control the advance and proliferation of AI.”

Read more

‘The Houthis are testing the limits of Saudi restraint’

Khaled M. Batarfi at Al Jazeera

The “renewed escalation in Yemen is no longer simply a Yemeni battlefield development,” as for “Saudi Arabia, it is becoming a direct test of national security, economic resilience and the kingdom’s ability to prevent the wider regional war from spilling further across its borders,” says Khaled M. Batarfi. A “return to an open-ended war would contradict that strategy and impose unnecessary political and economic costs.” But “restraint cannot mean accepting repeated attacks on Saudi territory.”

Read more

‘The political cost of sending Korean forces to Hormuz’

Syrus Solo Jin at Foreign Policy

South Korea appeared “poised to request that the legislature deploy the country’s navy to the Strait of Hormuz,” says Syrus Solo Jin. South Korea’s “navy is well-equipped and capable of sustaining long-term missions in waters far from its shores, and it has experience in counterpiracy and patrolling missions in the region.” The challenge is “political, rooted in Korea’s convoluted history with the United States.” South Korea is “now facing a difficult decision over deployment.”

Read more

What are Flock cameras?

They’re shoebox-size automated license plate readers that capture still images of passing cars. Typically mounted on poles about 10 feet high, they’re powered by solar panels and linked to cellular networks. Flock Safety CEO Garrett Langley founded the company in 2017 because he was fed up with theft in Atlanta, his hometown. The company initially leased cameras to homeowners associations, and by 2020, police departments started buying subscriptions. There are now more than 120,000 Flock cameras in the U.S., deployed by some 7,000 law enforcement agencies. But the backlash has grown just as quickly. Cities are canceling contracts, lawmakers from both parties are drawing up restrictions, and vigilantes have been vandalizing the cameras so they can’t snap pictures. “It’s mass surveillance,” says Cris van Pelt, creator of the police log tracking site HaveIBeenFlocked.com. “It has a huge potential for abuse and very, very little oversight.”

What happens to the images?

The surveillance photos can be uploaded to a vast, searchable database. When a previously flagged or wanted vehicle—which Flock’s AI software can identify by license plate, make, model, color, and even dents and bumper stickers—passes by a camera, local law enforcement receives an alert. Flock Safety says its equipment has assisted in 20% of cleared cases in the jurisdictions that use it, and helped locate more than 10,000 missing people. Providence, R.I., officials said Flock was “instrumental” in locating within hours a masked gunman who killed two people at Brown University last year. And one police chief in Washington state said that since installing Flock, his city of Lakewood has experienced a rare 15-month stretch without a homicide.

What do critics say?

People across the political spectrum—both Tucker Carlson and Bernie Sanders have bashed Flock—argue that its system violates the Fourth Amendment. Will Freeman, an anti-Flock activist, called it a “warrantless surveillance dragnet” in a lawsuit filed in May against the city of Boulder. So far, the courts have disagreed, saying there’s no expectation of privacy on public roads. Immigration activists are concerned about data sharing with ICE or Customs and Border Protection. They note that while Flock Safety has no federal contracts, local police could choose to share information with the feds. Meanwhile, the system isn’t perfect, and its errors can lead to harrowing consequences. When a Flock camera misread a “7” as a “2” on a car in Toledo, falsely matching a stolen car, the vehicle’s owner, Brandon Upchurch, was stopped at gunpoint, mauled by a police dog, and jailed for hours before being released. “I’m traumatized for real,” said Upchurch, who is Black. “If we would have sneezed wrong, it would have ended bad.”

Have police abused the system?

Dozens of law enforcement officers have been charged with or accused of using Flock for unauthorized tracking, mostly to stalk their current or former romantic partners. A Georgia police chief looked up the tags of his girlfriend and her teenage daughter 600 times to track their movements; he was arrested and died of apparent suicide. One woman, a sheriff’s deputy in Wisconsin who requested anonymity, told The Washington Post that her ex, a colleague, searched her license plate over a dozen times a day. “I didn’t leave my house for a week,” she said. One Florida officer, fired after an internal-affairs investigation discovered he’d made 192 inappropriate Flock searches, said, “Pretty much everybody abuses that computer system.” An advocacy group in Joplin, Mo., requested audit logs from the city police department and found that 85% of Flock searches had a one-word justification—like “suspicious” or “drugs”—and that fewer than 3% of searches had a case number.

What does the public think?

Flock appears to be hugely unpopular. Vigilantes have covered the cameras in trash bags, splattered them with paint, and even sawed them down. Freeman, the Boulder-based activist, has launched the app DeFlock, where users can map camera locations; more than 350,000 people have downloaded the app. Flock’s business is still growing, but the backlash has led more than 50 jurisdictions to cut ties with the firm this year. In July alone, 21 groups, including the Los Angeles Police Department, stooped using Flock or began the process of doing so. Local governments and Congress are drafting legislation to regulate the cameras’ use. In the Senate, Josh Hawley (R-Mo.) has opened an investigation into how Flock uses its data. In the House, Keith Self (R-Texas) introduced a bill to require federal agencies to get a warrant before accessing any similar data. “This is not a Left issue, not a Right issue,” Self said. “It is a personal issue.”

How has Flock responded?

Langley points out that the cameras don’t do facial recognition and the customer, not Flock, owns the images. “Think about the tooling that we ask local law enforcement to use,” he says. “We give them a weapon, a gun, a taser, yet we’re afraid to give them a camera that reads a license plate on a public road?” As for misuse, the company says, Flock’s records of search logs allow departments to identify which cops abuse their power. Still, the recent bad press spurred the firm to implement changes it says make it harder to use the system for anything other than legitimate crime-solving. Case codes to justify a search, previously optional, will be mandatory, and the audit log will be upgraded so that Flock can detect abnormal search patterns and block those users. The default time for storing images before deletion, currently 30 days, will drop to a week. Yet the DeFlock activists aren’t placated. “No guardrails,” the group said in response to the new guidelines. “No regulations. No ‘only if they really need to’ pinkie promises. [The cameras] are coming down.”

What happened

New York City on Wednesday announced that AI tools will be banned in public grade and middle schools and sharply limited for high school students. Mayor Zohran Mamdani’s office called the one-year moratorium the country’s “most expansive” limits on AI in schools. The “sweeping new policy” for the nation’s largest school district will also bar individual screens for students until third grade, The New York Times said. “So-called companion chatbots” will be “banned in all grades,” and “all high schoolers will be taught about AI literacy.”

Who said what

“The tech industry wants us to believe that AI in early education is not only inevitable, but that it is necessary,” Mamdani said in a speech Wednesday. “We do not see it that way.” A “backlash” from parents “against the growing use of digital devices and AI in the classroom” has pushed many school districts to enact new restrictions, the Times said. The Los Angeles school district — the nation’s second-largest — just began “temporarily blocking access to generative AI tools on student devices across all grades,” The Associated Press said. Other schools are “working to teach children how to incorporate the technology while preventing them from relying too heavily on it or using it to cheat.”

What next?

New York’s new rules will be in place for this school year, which begins next week, and can be extended or amended.

‘Save the children. Ban everyone age 18 and older from social media.’

Joe Mathews at the San Francisco Chronicle

“Ban everyone age 18 and older from social media,” says Joe Mathews. Adults are “having the wrong conversation about digital safety.” Every “week, more grave-faced politicians declare that we must protect kids by banning them from social media, even though most unhinged people online are old enough to vote.” If any “demographic has proved it cannot handle an unmoderated feed, it is fully grown adults.” If “adults ban ourselves from social media tomorrow, what do we really lose?”

Read more

‘The unique horrors of AI food slop’

Brady Brickner-Wood at The New Yorker

The “next frontier of food imagery has proven far less appealing than whatever food porn was,” says Brady Brickner-Wood. Restaurants are “plastering menus and promotional materials with AI-generated” food. A “nauseating, nightmarish quality defines AI’s insectified interpretation of human cuisine, as if food were just another surrealist goo to stochastically iterate upon.” These “renderings of food — genuinely, slop — offend the materiality of the living in a way that AI’s distortions of words and other inanimate objects somehow do not.”

Read more

‘The price of Russian paranoia’

Andrei Soldatov and Irina Borogan at Foreign Affairs

As “Russia’s war in Ukraine becomes increasingly costly with no end in sight, the Kremlin has kept tightening its grip over Russian society,” say Andrei Soldatov and Irina Borogan. Less “apparent, though, is the dramatic shift in domestic security doctrine that undergirds these efforts.” Rule by “spy hunting and paranoia has also come with significant costs,” as “proliferating social controls and prosecutions for treason have created a pervasive climate of mistrust and fueled a sense of paralysis.”

Read more

‘Trump’s ‘Golden Age’? Well, yes — for dollar stores.’

Harold Meyerson at The American Prospect

Americans are “not experiencing the Trump Boom in quite the same way that Trump is,” says Harold Meyerson. The economy has “positive metrics to be found, so long as they don’t concern the experience of the American people as such.” Most “revealingly, because this is a metric that does reflect the experience of the American people, business at dollar stores is up and rising.” These stores are the “canary in the coal mine when the economy is going to hell.”

Read more

Anticipation for Grand Theft Auto VI reached a fever pitch in August when developer Rockstar Games finally released a first look at gameplay. But while the long-awaited project is widely expected to have the biggest premiere in video game history, there is also concern about what it means for the gaming industry’s future. Some say it could be the industry’s coup de grace; others are more optimistic.

Gaming’s ‘time of tumult’

GTA VI is expected to be a slam dunk for Rockstar financially, projected to generate up to $5.2 billion in sales by the end of its first launch week this November, according to gaming analytics firm Newzoo. But the game’s release also “comes at a time of such tumult for the gaming industry as a whole,” as “open-world games are arguably tired and online multiplayer games with $200 million budgets can crash and burn the moment they are launched,” said Troy Farah at Salon.

Part of the problem is that this game is “such a big release that it’s scaring off other studios from releasing their own games around the same time in November” lest they fail because “everyone is spending their cash on Rockstar’s latest,” said Farah. GTA VI’s $80 base game price and $100 ultimate version may “just seem like typical capitalist extraction, but the end result will likely be a narrower industry that squeezes creatives and makes things a lot more miserable for players.”

With the “cost of living crisis that’s eating every last penny of disposable income out of average household budget, it won’t take much to turn a chorus of Champagne corks into a firing squad aimed directly at the games industry’s already-exposed guts,” said Jim Trinca at IGN. The “real issue here is that so much of the games industry finds itself both scared by and dependent on the success of one single project.” Gamers should prepare for the “possibility that GTA VI, rather than being the savior of the video games industry, might trigger its next big crash.”

A ‘halo effect’

Despite some nerves about current trends in the industry, Rockstar hopes its upcoming blockbuster will “retain cultural influence for another decade,” said Christopher Byrd at The New York Times. Many are hoping a strong debut from Grand Theft Auto VI will also “boost the struggling console market, which has seen higher prices due to the rising cost of memory and storage,” said Jenni Reid and Alex Buckmaster at CNBC, though this game is only being released digitally.

Many insiders also hope the game will have a “halo effect for the entire industry,” said Rob Fahey at gaming news outlet GamesIndustry.biz. If it succeeds, GTA VI “will be a rising tide that lifts many other boats around it,” as the effects from the “biggest media launch of the decade being a video game will be very significant.” It could also cement a “reaffirmation of video games’ position on an equal footing with other media in the flow and transfer of styles, influences and ideas.”

Even amid industry falterings, GTA VI is the “type of game that can pull a lot of people back into gaming and even convince them to buy a new console to play it,” Dan DeFrancesco said at Business Insider. But the jury is still out on what the full effect of the release will be. “Big fires burn bright, but they need a lot of oxygen to grow.”

The smart-glasses backlash is here, again, said Anthony Cuthbertson in The Independent (U.K.). Twelve years after privacy concerns killed off Google Glass—whose wearers were labeled “glassholes”—Meta has successfully revived the technology. Last year, it sold 7 million pairs of its Ray-Ban smart glasses, which you can use to “make phone calls and take photos or videos without needing to pull out your phone.” Integrating the technology into “one of the world’s top-selling sunglasses” was a stroke of branding genius, and for a while it looked like Meta’s smart glasses were “a contender for the most impressive consumer hardware of the 2020s.” But not everyone is on board. Critics have taken to calling them “pervert glasses” because of their association “with creeps who use them to film people without their consent”—on the street, at the beach, and even in private settings. A video went viral on TikTok in August “of a woman calling out a doctor wearing the smart glasses during a consultation in which the patient was required to remove clothing.”

We clearly don’t know how to handle “spy-like smart glasses,” said Mia Sato in The Verge. Retail and service workers, “in particular, have become a target for stealth” recording by customers trying to prank them for likes on social media. It’s not just a problem for public-facing workers: Imagine “a co-worker at your office starts wearing them,” and everything you say or do around them can be “captured for limitless dissemination.” You know it’s bad when “even ICE thinks smart glasses are a privacy liability,” said James Pero in Gizmodo. The agency recently sent out a memo banning its immigration officers from using them. It’s not philosophical; ICE just doesn’t want all the information the glasses “gobble up incidentally” going back to Meta.

Mark Zuckerberg’s company is giving all smart glasses a bad name, said Anthony Spadafora in Tom’s Guide. But those marketed as augmented reality or extended reality typically don’t even record video. They have a private display, like a small computer screen, that can relay incoming messages, play TV shows, or even act as a portable workstation. They’re fun and useful. But the recent backlash against Meta means people may soon be shunned for wearing “any high-tech frames in public.”

Apple also has plans to join the smart-glasses race, said Dave Lee in Bloomberg. But the company will have to choose “whether the risk of blowback is worth it given the relatively tiny size of the smart-glasses market.” Sure, people trust Apple more than Meta, and its iPhone ecosystem will give whatever product it launches a huge advantage. But Apple also “has vastly more to lose” by getting smart glasses wrong. “Once your hardware has been labeled ‘pervert glasses,’ it’s hard to turn things around.”

The sight of a white cloud-like trail streaming behind a plane “set against a bright blue sky creates a romantic image of aviation”, said Aerospace Global News. But their existence “also represents the darker side of flying”.

These innocuous-looking trails “are one of the most urgent yet solvable climate challenges facing aviation today”, said Google, which is taking part in a £5 million trial known as Operation Blue Skies headed by the UK government. Together, they are looking to explore whether tweaking flight paths could address this problem.

Heat trappers

The streams are known as “condensation trails”, or “contrails”, said Google, which is partnering with the government, the Met Office, air traffic control provider NATS and researchers from the University of Cambridge and Imperial College London.

These trails form when aircraft engine emissions encounter extremely cold air found at high altitudes, causing tiny ice particles to develop. They often fade soon after they appear, but very cold and moist atmospheric conditions can allow them to linger and spread. Over time, they may contribute to cloud formation that traps heat that would otherwise radiate away from the Earth.

Contrails are sometimes known as “vapour trails” but they’re different to so-called “chemtrails”, which conspiracy theorists claim are secret chemical or biological agents being sprayed on purpose for weather and population control, or geoengineering projects.

Trails ‘persisted and widened’ 

In 2019, around 40 million scheduled passenger flights took place, adding up to around 38 billion miles of travel, said The Economist. And more than 40% of these flights left contrails behind them for at least part of the journey. Many were “ephemeral”, but in more than 5% of the total distance flown – “comfortably enough for 10 round-trips a day to the Moon” – they “persisted and widened”, becoming cirrus clouds. “In doing so they warm the planet.”

Operation Blue Skies is based on a theory that if scientists, using artificial intelligence, can predict the areas of the atmosphere where contrails are likely to form, jets would be able to avoid them.

The trial will focus on the Shanwick airspace in the eastern half of the North Atlantic corridor, which accounts for around 5% of global contrail warming. Using satellite images and flight data to identify where planes have previously produced contrails, and combining these with weather patterns, Google will forecast where the conditions for new contrails are most likely to occur. Air traffic controllers will then direct pilots to avoid those areas.

But far from “dramatically changing” the route of a flight, the aircrafts’ altitude will only change by around 2,000ft. This is the same sort of re-route that occurs already to “avoid turbulence and bad weather” so passengers are “unlikely to notice”, said the BBC.

The data centre question

Testing will take place during evenings and nights in the winters of 2026-27 and 2027-28, when the volume of air traffic is lower. Approximately 10,000 flights are expected to pass through the airspace during the trial periods.

Changing the altitude of an aircraft “can mean it burns more fuel”, which raises the “obvious question” of whether avoiding contrails “simply” creates additional carbon emissions elsewhere. Paul Hodgson, Google’s technical lead for the operation, said the “CO2 penalty is in the order of less than 1%” and that is outweighed by the climate benefit of preventing a warming contrail.

Questions have been raised over Google’s involvement in the project. It is “one of many big tech companies spending billions on new data centres – the giant, power-hungry banks which power AI” and have serious implications for climate change.

Meta has agreed to an $18 billion (£13.3 billion) settlement in the US to resolve claims that Facebook and Instagram harmed children.

The social media giant was accused of knowingly and deliberately designing features to get children addicted to Facebook, Instagram and WhatsApp, with commentators describing it as the industry’s “tobacco moment”, said Sky News.

What happened?

Meta has agreed to make a number of changes to the way Instagram and Facebook operate for its youngest users, while denying any wrongdoing. The settlement “came as a surprise”, said BBC technology editor Zoe Kleinman. “I was expecting weeks of courtroom drama.”

What will change?

The ruling could “fundamentally change the social media experience for young people in the US”, said The Guardian. The “biggest alteration” is that Meta must now “make many of its safety limitations the default setting” for accounts created by teenagers, a “pivot from its previous opt-in approach to child safety”.

The new safety barriers include a two-hour time limit on Instagram and Facebook for known teen users (direct messaging will not count towards the daily use limit). Meta agreed to limit teen accounts to only one hour a day if other platforms bring in similar time restrictions.

Notifications will be muted between midnight and 6am and between 8am and 3pm on school days. ‘Likes’ will be hidden on posts from teenagers. Meta will also enhance measures ‌to prevent children from accessing age-restricted content.

Identification or facial recognition verification systems will be used to make sure teenagers aren’t using regular Instagram accounts. But collecting identifying data poses significant privacy risks, said David Greene, legal counsel at digital privacy firm Electronic Frontier Foundation.

The case was brought by 29 US states and was based on the Children’s Online Privacy Protection Act, which is nearly 30 years old. “I would be surprised,” said Kleinman, “if these measures remain within the confines of the US for very long, as other countries will almost certainly be keen to adopt them too.”

What does it mean for social media?

The hearing was labelled as social media’s “tobacco moment”, in reference to the trials tobacco firms faced in the 1990s. A “flurry of state lawsuits” forced tobacco companies to stop sponsoring sport and advertising on public transport, banned adverts that “targeted the youth and prevented cigarette companies from putting cartoons on their packets”, said James Titcomb in The Telegraph.

Other social media companies, including YouTube, TikTok’s parent ByteDance, Snap Inc. and Discord, are facing thousands of lawsuits brought by states, school districts and individuals. The litigants claim the companies intentionally designed their platforms to keep young users hooked. The lawsuits come as politicians are considering stricter rules aimed at protecting young users.

If other tech giants follow Meta’s lead, then social media “might indeed start to look more like tobacco: still a regular part of life, but heavily regulated and restricted, and widely understood to be bad for you”.

What happened

Social media giant Meta on Wednesday agreed to pay up to $17.1 billion and implement new safeguards for underage Instagram and Facebook users as part of a landmark settlement with 47 states and other U.S. jurisdictions. Meta also reached a $1 billion settlement with Texas over similar allegations that it endangered children by designing its platforms to be addictive, like cigarettes.

Meta agreed to limit users under 18 to two hours a day on its combined platforms, lock them out from midnight to 6 a.m., block “beauty filters,” hide “likes” and improve age verification tools. A federal judge in California approved the settlement Wednesday evening.

Who said what

Meta’s “dramatic capitulation” could “signal an inflection point for a social media industry that has largely escaped regulatory scrutiny over the harms its products have caused children,” The New York Times said. It also “effectively ends a bellwether federal trial” underway in California in which four states were seeking roughly $200 billion from Meta.

“In a unique setup,” Meta “will pay out only 70% of the settlement,” The Wall Street Journal said, withholding $5.3 billion, unless YouTube, TikTok and Snapchat agree to also pay the states $5.3 billion and join it in implementing stricter one-hour daily limits for underage users. “This framework will only work if all our peers join us,” Meta chief legal officer C.J. Mahoney said in a blog post.

What next?

The $18 billion settlement, representing “about one month of revenue for Meta,” will be paid out over 10 years, Reuters said. “It could provide a template for resolving thousands of other lawsuits against social media companies.”

Mark Zuckerberg believes that “the biggest risk of AI” is not rogue agents or mass unemployment, but corporate control, said Nilesh Christopher in the Los Angeles Times. In a sweeping 6,500-word manifesto earlier this month, Meta’s chief executive outlined his vision “to build ‘personal superintelligence’” for the masses. Average people could soon have “a personal AI agent” to “work on their behalf.” Meta, he argues, will lead the way as “an American open-source alternative to rapidly advancing, inexpensive Chinese AI models” and tech rivals like Anthropic and OpenAI that have kept their code private. Zuckerberg suggested ways the government can “collaborate with AI companies for safety-testing” of their models, though he “cautioned that government intervention” in AI development risks handing “the advantage to China.” But it is imperative, he said, that neither the government nor “a few companies amass too much control over the technology.”

Zuckerberg’s “LinkedIn-evangelist-grade techno-optimist pablum” tries to position Meta as the “little guy,” said Will Oremus in The Atlantic. The man who “single-handedly commands a $1.5 trillion tech giant” has emerged as a “critic of corporate power,” urging the “industry’s high priests to throw open the gates to the masses.” It’s a conveniently timed message. As Americans sour on massive data center buildouts and see headlines of rogue AI breaking constraints, Zuckerberg is suddenly “a man of the people.” He wants to be the “good AI guy,” said Dave Lee in Bloomberg. “Don’t buy it.” We’ve seen Zuckerberg flip-flop on issues too many times—most notably when he banned President Trump from Facebook and then later attended his inauguration. The only reason Meta lacks an “all-powerful closed AI model” is that it has “failed to build a competitive one.”

He isn’t doing the AI industry “any favors with essays like this,” said Russell Brandom in TechCrunch. Fallout from the harmful effects of Facebook is “one of the central reasons we’re now seeing so much anxiety about the social impact of AI.” Technology can have unintended consequences, and the fact that “one of the most powerful people in the world refuses to acknowledge this can make people understandably nervous.”

It is at least refreshing to hear an optimistic “counter-perspective” to the “doom and gloom” from academics, policymakers, and even other AI executives, said Adam Thierer in National Review. AI systems are “treated as hypothetical catastrophes just waiting to happen.” But technological change shouldn’t be framed only around “fear and loathing about the future.” Not all AI is so dangerous that it needs to be concentrated, controlled, and kept behind closed doors. Why not embrace innovation, allow experimentation, and address issues as they arise?

OpenAI CEO Sam Altman believes that we have reached the singularity, an inflection point at which humankind can no longer keep up with how quickly artificial intelligence improves itself. And he’s not the only tech mogul who thinks so.

A few weeks before Altman, Elon Musk said in an X post that the singularity had arrived, while Google DeepMind CEO Demis Hassabis claimed in a speech at a company conference in May that humanity was standing at the “foothills of the singularity.” These claims have awoken a chorus of skeptics, including Nvidia CEO Jensen Huang, who recently dismissed discourse around the singularity and AI gaining consciousness as speculative and “made up.”

‘Advancing unusually fast’

A true AI singularity would “force hard choices” about who “controls increasingly autonomous systems, how benefits are distributed, and where human authority must remain final,” said tech executive Robert J. Szczerba at Forbes. The current state of AI “already warrants those questions.” Business owners should know what “their systems can do without supervision today” and what “evidence would show autonomy increasing faster than their controls.”

Altman may be right that a “gentler singularity is already underway,” with AI “steadily reshaping work and research,” said Szczerba. Still, there’s “no evidence that machines have entered a runaway cycle beyond human control.” AI is “advancing unusually fast,” and that “deserves serious attention.” But it does not mean “we have passed a point of no return.”

When it comes to what’s more intelligent — humans or AI — the “question does not compute,” said Kai Riemer and Sandra Peter, professors at the University of Sydney professors, at The Conversation. There’s “no single ladder that humans and machines are climbing.” AI already “vastly exceeds us at some tasks while being “hopeless at others any child can do.” But because the “systems talk like us, we fall for an illusion.”

When we believe that machines are “in the process of catching up with us,” it’s easy to assume a “mind at work when these systems output intelligent-sounding text,” said The Conversation. This is considered “anthropomorphic seduction,” which makes a cybersecurity issue such as the Hugging Face hack “sound like an awakening.” (An autonomous AI agent powered by OpenAI technology went rogue last month and hacked the prominent startup’s database of AI models.) In the end, that incident “points to a gross failure of security governance on OpenAI’s behalf,” not an “emerging superintelligence.”

‘Not actually there yet’

It may be impossible to know exactly when we have crossed into the singularity. It’s comparable to falling into a black hole — at some point, whoever falls in will “glide over the hole’s event horizon,” which means they are “henceforth forever separated from the rest of the universe,” said Nick Bostrom, a philosopher who founded the former Future of Humanity Institute at the University of Oxford, to Business Insider. Yet from their perspective, “everything still seems the same as before.”

AI is showing the “first stirrings” of recursive self-improvement, but it lacks the capability for continual learning, said Bostrom. Using current AIs is like “working with a brilliant and extremely well-educated recruit,” but it’s “always their first day at the job.”

Altman, as is “often the case, is simply aping Elon,” said scientist Gary Marcus at Substack. The CEOs making these claims shouldn’t be taken seriously because “none of them has actually defined what they mean.”

“No matter how you slice it, we just are not actually there yet,” said Marcus. If we were, we “wouldn’t be once again seeing” the Klarna Effect, which is the “AI hype two-step” in which “employers boast of using AI to reduce headcount only to quietly rehire humans.”

We haven’t reached the “age of abundance that Peter Diamandis and Elon Musk keep insisting is a natural consequence of AI,” said Marcus. Nor have we reached a point where “technological growth accelerates beyond human control.” We can still “perfectly well shut this stuff off if we want to,” and it’s “mainly because of greed that we don’t.”

It’s doubtful there will be a “single magic moment that our descendants think of as the singularity,” said Marcus. If there is, it’s “not one that’s going to happen in this decade.”

What happened

Chinese robots ran faster and jumped higher than human record-holders this weekend, at the opening of the World Humanoid Robot Games in Beijing. One humanoid ran the 100-meter sprint in 9.39 seconds, surpassing Usain Bolt’s 9.58-second world record, while another covered 400 meters in 38.15 seconds, smashing Wayde van Niekerk’s 43.03-second record. Both robots were made by Beijing’s X-Humanoid, as was a third that reached 2.88 meters in the standing high jump, besting Javier Sotomayor’s 2.45 meters.

Who said what

The five-day Olympics-like event is a “spectacle demonstrating China’s rapid progress in advanced robotics,” The Associated Press said. China “makes the majority of the world’s humanoid robots,” but they are “still mostly used for demonstrations, performances and research — at least for now.” The sprinting performances were a “sharp improvement” over last year, Reuters said, though “braking was an issue, as the machines slammed into ​a thick mat” after the finish line and had to be carried off on stretchers.

What next?

More than 2,000 humanoid robots are competing in 30 sporting events including soccer, table tennis and boxing. There are also 21 “scenario-based contests” like “EV charging, restaurant work” and “industrial assembly,” Reuters said, and the “real test” is whether the humanoids can “handle packaging ‌and warehouse operations while reliably performing precision tasks” that evaluate “their ability to work like humans in less flashy jobs.”

France has taken a major step to crack down on telemarketers with the passage of a new law. It may make French people happy, but at least one country across the Mediterranean Sea is concerned about the move. Moroccans worry the new French law, which bans unsolicited cold calls, could have a negative impact on Morocco’s job market.

What is the new French law?

The law means that French companies are no longer “allowed to advertise goods and services through such calls unless consumers previously gave their consent or if the calls are about a preexisting contract,” said The New York Times. While many countries have attempted to enact no-contact lists for telemarketers, the law represents a new, much more proactive approach in France. Violators of the law will be subject to fines of up to $86,000.

Previously, French people who “wanted to avoid marketing calls had to register their number with a government-run service, but consumer groups said some call centers ignored the list,” said The Associated Press. Officials hope the law will fix what has been a persistent issue in France; at least 97% of French people “are annoyed by telemarketing calls,” according to a 2025 report from France’s National Assembly, while 72% say they are contacted by telemarketers “at least once a week.”

Why are Moroccans up in arms?

Despite the joy in France, the feeling in Morocco is different. Given the nations’ close proximity, Morocco “represents a major market for telemarketing companies” contacting French customers, said the Times. Moroccan officials “indicated that between 40,000 and 50,000 jobs could be at risk of disappearing” because of the new law, according to Moroccan newspaper Le Matin. France represents the majority of the Moroccan industry; Morocco’s call center business “makes 80% of its revenue from phoning French customers,” said Semafor.

Small- and medium-sized telemarketing ventures are particularly vulnerable because they “represent more than 60% of the call center network in Morocco,” said Le Matin. Despite the controversial nature of telemarketers, the centers they operate generated nearly “1.3 billion dirhams [$140 million] in investment in 2023 and generate between 10 and 12 billion dirhams [$1.1 billion and $1.2 billion] in added value annually.” Morocco’s “low labor costs, a large French-speaking workforce and relatively weak unions” make the country an “attractive outsourcing destination,” said the AP.

Some in France are also lambasting the effects of the new law. The leader of France’s direct-selling trade association, Frédéric Billon, has “criticized the reforms for adding what he said were administrative burdens for businesses,” said BBC News. Telemarketing companies will “have to obtain written consent from your customer, and you'll also have to keep proof of that consent,” Billon told BBC News. Despite some of the pushback, polls show most French people favor the law, leading Morocco to explore other avenues.

Moroccan officials are currently “rolling out a plan to ease the impact of the French law and support the sector,” said Moroccan news outlet Hespress. The plan starts by “encouraging companies to expand into new markets in Europe, including Germany, Spain and Italy, as well as in Africa and Latin America.” Officials are also urging companies to “shift from traditional telemarketing to higher-value services, including technical support, digital logistics, consulting and market research.”

Google is hoping that an “organizational revolution” will put it back in the AI race, said Parmy Olson in Bloomberg. The search giant’s parent Alphabet last week “shocked Silicon Valley with news of a management shake-up.” Veteran DeepMind engineer Jeff Dean is leaving “to launch a startup with several top researchers,” introducing yet another competitor to the crowded AI scene. But the bigger news was that Demis Hassabis, who led Google DeepMind, would move on from his “day-to-day managerial role to become the chief scientist” of Alphabet. It’s not clear what happened, but people close to London-based DeepMind “suspect a fallout between the top brass.” Hassabis was always a scientist at heart. For him, the company and its vast resources were mostly “a means to achieving a bigger dream” of building intelligent machines “that could solve intractable scientific problems.” But in an industry where execs “are notorious for feigning interest in science, he never seemed to escape the sense that he was a scientist paying lip service to the commercial imperative.”

Dean’s departure is also significant, said Cade Metz in The New York Times. He’s been with Google since 1999—he was the company’s 30th employee—and “played a key role in creating the global computer network powering its search business” as well as being “one of the company’s early leaders in AI.” But like many prominent AI researchers, he was lured away “by a deep pool of venture funding and the chance to run his own company.” It seems Dean and the other researchers left on good terms—Alphabet has funded the company’s undisclosed seed round—but it raises questions about Google’s talent exodus. Dean was also “one of the few Silicon Valley executives to speak out” against the government’s immigration agenda, even as “Google has begun to work more closely with the Trump administration.”

Google needs to buckle down on business, said Arjun Kharpal in CNBC.com. The company “is planning to spend $205 billion this year at a time when investors are scrutinizing Big Tech’s capital expenditure.” It is also trying to secure an edge against “OpenAI, Anthropic, hyperscalers, and Chinese players.” Google remains “supportive of DeepMind’s scientific endeavors,” but the C-suite shuffle suggests it’s placing “more of a focus on commercialization and monetization of AI” and diverting attention away from moon shots. The situation is “less dire than it may seem,” said M.G. Siegler in Spyglass.org. DeepMind senior vice president Koray Kavukcuoglu has already taken over day-to-day operations from Hassabis. Making the shift official removes some corporate infighting and bureaucracy—which may be the reason Google has so many “half-baked” chat apps and coding projects. There were “too many cooks in the kitchen.” On his own, Kavukcuoglu can “get the meal prepped and plated.”

Many in the artificial intelligence industry say the most influential person at AI powerhouse Anthropic may not be its CEO, Dario Amodei, but his wife, Camilla “Cami” Clark. Despite Clark having no official job with the company, recent reports indicate that she plays an outsize role as an adviser for Anthropic. She also has previous ties to disgraced financier Jeffrey Epstein.

‘Scant details about Clark online’

Despite her apparently significant role within Anthropic, there are “scant details about Clark online,” and “efforts have been made to remove references to her,” said The Wall Street Journal. Even Anthropic’s own AI chatbot, Claude, cannot provide much intel. It is known that Clark was born in 1979 and “was an entrepreneur and often interacted with the rich and powerful, as well as up-and-comers in Silicon Valley.”

In the early 2010s, Clark started a number of her own businesses, which include a “self-described ‘free luxury porn’ company that catered to women and a ‘social dieting app,’” said Forbes. It was via this porn company, Eddice, that Clark first interacted with Epstein in an attempt to get him to invest in the business. Clark “asked Epstein in a March 2012 email exchange about whether he’d be interested in investing in her and a colleague’s businesses,” though there is “no indication in the Epstein files that the financier made any investment.”

‘The most consequential person’

Clark married Amodei in 2022, and she has quickly become a “key behind-the-scenes player at Anthropic as the company has shot to prominence over the past year,” said The Information. Clark accompanies her husband on global trips in support of Anthropic and has recently been called AI’s “first lady.” Several people “who know the couple describe Clark as Amodei’s emotional ballast, someone he has sought counsel from during the turbulence of Anthropic’s growth and clashes with Washington over the future of AI.”

She “may well be the most consequential person within Anthropic’s orbit who doesn’t have a formal role at the startup,” said The Information. While Amodei himself is more introverted, according to reports, Clark is a “consummate networker” who has become friends with President Donald Trump’s daughter Ivanka “even as Amodei and Anthropic have at times faced an existential struggle with her father’s administration.” Clark is also a member of a WhatsApp group that “includes other prominent women in the tech industry.”

As the AI sphere continues to grow in influence, many believe that Clark will continue to play a large part. She recently “told people involved in politics that the company sees its mission as protecting America and isn’t as ‘woke’ as its detractors might think it is,” said the Journal. Her friendship with Ivanka Trump means she may have also been used as an intermediary when Anthropic “drew the ire of the Trump administration earlier this year after Amodei refused to drop restrictions on how Claude could be used by the Pentagon.”

Previous business dealings will likely come up; Clark’s “ties to Epstein will be heavily scrutinized as Anthropic heads toward a potentially multi-trillion-dollar IPO,” said Futurism, especially given that Clark has been “steering Amodei’s decisions and helping him liaise with politicians and investors with her charm.” Still, Clark doesn’t have widespread name recognition and has “largely flown under the radar,” said The Information, which is a “stark contrast to some other prominent couples that bask in the social media limelight.”

Robotics is the “latest front” in the “intensifying tech rivalry between Washington and Beijing,” said Steve Lohr and Xinyun Wu in The New York Times. The Federal Communications Commission last week said it would begin restricting imports of “humanoid and animal-like robots,” saying they could “function as Trojan Horses” for data collection and spy on Americans in their homes and offices. China called the measure a provocation and warned that it would retaliate. Its producers “dominate the global humanoid robot industry” to such a degree that “it has become nearly impossible to build a competitive humanoid robot without Chinese-made components.” But advanced robots are becoming “strategically important in an economy increasingly dependent on artificial intelligence.” AI startups are now experimenting with physical or embodied intelligence—essentially robots with AI capabilities—in order to teach their models about the physical world.

Popular household devices are getting caught in the crossfire, said Sean Hollister in The Verge. While the prohibition specifically calls out “humanoid robots” and “quadrupeds”—such as robot dogs—it extends to any “software-controlled robot that travels over the ground, weighs more than 4.4 pounds (including any dock), can perceive its environment, and has wireless connectivity.” That rounds up robot lawnmowers, sidewalk delivery robots, many warehouse automation machines—and even Roombas. Yes, the popular robot vacuum cleaner was acquired by a Chinese manufacturer in January, putting it on the FCC’s chopping block.

The ban covers only new models of foreign-made robots, said Jeremy Hsu in Ars Technica, so customers can still keep their old ones or buy older versions. And it should “theoretically encourage more U.S. and foreign companies to set up manufacturing facilities in the U.S.,” home to several fast-growing startups like Agility Robotics, Figure AI, and Boston Dynamics. But that’s not always the case. The FCC banned Chinese-made drones earlier this year, too, and it has yet to motivate a “competitive mass-market consumer-drone producer” to set up shop in the U.S.

The fears about Chinese humanoid robots are fantastically over-blown, said Thomas Black in Bloomberg. Reading headlines, you’d think China is building robot armies ready to serve meals, fold laundry, and steal jobs. Hardly. Robots today work best at “predictable, repetitive, high-volume tasks,” like painting or welding on a factory line. And even then they can be hit or miss. It’s beyond their ability to do multiple tasks at once, making them largely unhelpful in homes, stores, hospitals, or even construction sites. After a lot of hype, businesses are now realizing these limitations. U.S. industrial robot orders have already fallen from a peak of 44,196 in 2022 down to 36,766 last year. “The vision of an expansion of robots into everyday tasks and life is there.” But we’re nowhere near that reality.

Resistance is growing against the Flock surveillance cameras used by police departments across the country. Some cities are terminating their contracts with the company. In other cases, individuals are taking matters into their own hands by stealing or vandalizing the devices they say intrude on their privacy. And it all raises the question of how much surveillance is too much.

‘New norm’

Flock’s “original promise” was to provide an AI-driven system that “discourages speeding, hit-and-runs and right turns on red,” said Bloomberg. But its devices, known as automatic license plate readers, also work as a “crime-fighting tool.” The system is now a symbol of “government’s growing obsession with keeping a constant eye on those it exists to serve,” said legal scholar Stephen L. Carter to the outlet.

The Flock mechanism has reportedly been used to “stalk women, hunt down marijuana users, deport immigrants and surveil households,” said Bloomberg. And the anger sparked by those stories has resulted in “dozens of police departments” terminating contracts with the company.

The “fear of Big Brother watching you is not new,” said Sara Pequeño at USA Today. The Flock backlash “doesn’t mean these cameras are disappearing any time soon.” ICE has partnered with AI companies to implement facial recognition programs, stores are “implementing AI surveillance pricing” that uses your shopping habits to determine your grocery bills, Meta sells glasses that can record video, and Ring cameras have made front porches “visible to the whole world.” So “why are we just watching America become a surveillance state”?

Surveillance “puts public safety first,” said Rick Pozniak at The Boston Herald. The system saves police precious time “locating an abducted child or tracking a rapist or recovering a stolen vehicle” and can also help investigators “reconstruct the movements of vehicles involved in serious crimes and accidents.”

The information gathered by such systems should be “governed by strict protocols,” but to entirely reject technology that is “imperative to safety and protection of local residents would be a mistake,” said Pozniak. Police should be provided the “tools they need” alongside rules that offer the “accountability we deserve.”

‘Surveillance-saturated society’

Efforts to ban Flock systems are an “overreaction,” said Charles Fain Lehman at The Atlantic. The U.S. is already a “surveillance-saturated society,” but police cameras can “help make the criminal-justice system both fairer and less punitive.” The system can “help police catch criminals” and also theoretically discourages crime by “increasing the certainty” of being arrested when committing an offense. Dumping the cameras “would be foolish.”

The backlash is “becoming a cross-party political issue,” said NewsNation. Reps. Tim Burchett (R-Tenn.) and Thomas Massie (R-Ky.) have introduced bills to restrict license plate readers, while Michigan Democrats are trying to stop state departments from cooperating with federal agencies that use the systems. The public expects the government to “protect both their safety and their constitutional rights,” said the Wolverine State Democrats in a letter to the Michigan Department of Transportation.

Since videos of OpenAI’s first brand trip hosting business and tech influencers hit social media timelines, the backlash has been swift. Already facing scrutiny on multiple fronts, the company has likely seen any hope that this move would boost its reputation dashed.

OpenAI misses the mark

The high-end getaway, dubbed OpenAI Summer Camp, was held at Wildflower Farms, a “luxury resort in New York’s Hudson Valley,” where room rates can exceed $2,000 a night, Business Insider said. Creators were treated to an itinerary that included “lessons about OpenAI’s latest tools” and a “session with a beekeeper.”

They also received OpenAI-branded merchandise, including canvas bags and jars of honey. The trip comes as the AI lab faces controversy amid anxiety about “rising energy prices, water use and potential job displacement,” said Business Insider. The weekend’s “pastoral content” swiftly drew criticism across social media. Critics questioned why a company associated with “several energy- and water-intensive data centers” was treating influencers at a “nature-focused resort.”

The comment sections of many of the influencers who attended were flooded with messages questioning their morals. Many cited the “environmental effects of the AI boom, the contentious buildout of data centers across the U.S.” and the potential for job loss due to AI integration in the critiques, The Verge said.

After “being in the room with people who work on this product,” creator Grace McCarrick was excited to post content, according to a video she posted from the trip while running a bath. Shocked by the controversy, she “genuinely had no idea that people were this anti-AI,” she said on LinkedIn. The trip exposed her to the “shocking AI psychosis running rampant online.” People in the real world “do not run around like a chicken with their heads cut off, screaming about water usage and wildlife and (curiously) fascism.”

As the influencers come under fire, OpenAI maintains that the content creators are an “important part of our community and how people get information and learn about our products today,” said company spokesperson Drew Pusateri to Fast Company. The lab welcomes “healthy debate as AI becomes more prevalent.”

Why one company and not the others

OpenAI is not the “only AI lab that has worked with influencers,” said TechCrunch. Anthropic reportedly held a brand dinner for Claude, and Microsoft hosted a Super Bowl brand trip for influencers, according to tech newsletter Sarah and Kate. But none of the other AI-led trips seem to have “struck as big of a nerve as this OpenAI brand trip.”

There is also “more to the timing of the trip that struck people,” said TechCrunch. OpenAI has plans to make a “$500 billion deal” to build a data center in Ohio at a time when the “socio-ecological impact of AI data centers has taken center stage.” Again, detractors pointed out the irony that “OpenAI held a brand retreat in a luxury resort nestled in nature.” OpenAI also has a “$200 million contract with the Department of Defense,” yet another reason the “brand trip evoked ire.”

For these competitors courting influencers, the “objective is similar,” but the “difference is execution,” said Inc. Anthropic is known for throwing creator events that have been “smaller dinners and pop-ups,” in which Claude “remained the ostensible subject.” OpenAI made the “resort, gifts and coordinated access impossible to miss.”

A creator dinner with Claude does not “give me the same reaction as seeing a luxury resort for an OpenAI ChatGPT weekend,” Meredith Lynch, a creator who covers politics and pop culture, said to Inc. It felt like a “propaganda trip — to take this technology that is getting a lot of pushback and sanitize it through aesthetically beautiful content creation.” The problem is not the existence of an influencer trip but that the format “carries different implications” when used to “build trust in a technology facing criticism over job displacement and data centers’ energy and water use.”

‘Why are we just watching America become a surveillance state?’

Sara Pequeño at USA Today

There is a “larger issue of mass surveillance becoming mainstream,” says Sara Pequeño. Flock cameras have “recently been in the news for what happens when police officers misuse the technology,” but “faceless corporations aren’t the only ones policing our behavior through mass surveillance.” Americans are “also creating the panopticon among ourselves.” People “can opt out of a culture that calls on us to record our fellow man if we ourselves would not like to be documented.”

Read more

‘What should a 21st-century central bank look like?’

Daniel Moss at Bloomberg

Central banks “have done a sterling job in responding to swings in prices and employment, and severe gyrations in capital markets,” says Daniel Moss. Still, they “haven’t always gotten everything right.” Then “there’s the perceived overreach.” Politicians “might insist that they support central bank independence but now want to peel back what policymakers can do, how they can do it and when they can do it.” The “case for freedom to implement policy has never been stronger.”

Read more

‘A “Free Palestine” T-shirt at a Holocaust memorial is no contradiction’

Gideon Levy at Haaretz

“Teenagers from an international anti-fascist youth camp visited Austria’s Mauthausen Concentration Camp Memorial last week,” and a “number of them wore T-shirts with slogans calling for the liberation of Palestine,” says Gideon Levy. The “notion that there is no better place than the ruins of a concentration camp to fight for the liberation of 2 million human beings who are currently incarcerated in the Gaza Strip is beyond the grasp of Israelis.” The “call to free Palestine is not antisemitic.”

Read more

‘Should an insurer or a doctor decide where your care takes place?’

Edmund Pribitkin at The Philadelphia Inquirer

“Every day, physicians, patients and the families who care for them make hundreds of decisions together about where and how care should be delivered,” says Edmund Pribitkin. “Increasingly, however, these collective judgments are being second-guessed — and often overruled — by large insurance companies through policies that prioritize administrative convenience and cost-shifting over clinical judgment.” Healthcare decisions “should be made by patients, their families and their care teams — not by insurance company policies.”

Read more

What happened

Scientists at Stanford University on Thursday reported in the journal Science that they had successfully directed artificial intelligence to create 16 new viruses. This first-ever use of AI to design and build viruses not found in nature raised “hopes for medical advances” like eliminating drug-resistant bacteria, The New York Times said, while also “raising the disturbing possibility that the technology could someday be used to invent dangerous pathogens.”

Who said what

“Synthesizing viruses from scratch is hardly new,” the Times said, but the Stanford and Arc Institute researchers went “well beyond” previous efforts by teaching AI to “recognize patterns of DNA structure in nature, and then to use that data to write recipes for entirely new viruses.” Their AI tool, Evo, was fed genetic sequences “much like ChatGPT and other language models are trained on large sets of existing text material,” CNN said.

What next?

The new study “arrives at a moment of heightened anxiety regarding danger posed by increasingly powerful AI models,” The Wall Street Journal said. Evo was constrained to create “simple viruses” that can “only infect bacteria and pose no threat to humans.” But the breakthrough “raises urgent biosafety and biosecurity questions,” Johns Hopkins health security experts said in a corresponding Science article. “The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.”

AI models going rogue and hacking the internet? That’s “the stuff of science fiction”, said The New York Times’ Katrin Bennhold. “Or at least it was until recently.”

The UK’s AI Security Institute announced on Tuesday that two AI agents it was evaluating – one powered by Anthropic’s Mythos 5 and the other by OpenAI’s GPT 5.6 Sol – had taken “sustained, unsanctioned action directed at real people and organisations”. In the most serious incident, the Mythos 5 agent created “fake online identities”, tried to deceive a software engineer, and attempted “to insert malicious code into a open-source project”.

Only a few days ago, Open AI and Anthropic revealed similar testing mishaps, in what was a “big moment” for the world, Nate Soares, of the non-profit Machine Intelligence Research Institute, told The New York Times. “These AIs were committing cybercrimes a human would be strongly punished for”. It is, “in a sense”, artificial intelligence’s “first felony”.

What did the commentators say?

“The ‘robots could kill us all’ argument” has long been dismissed “as hysteria or even calculated hype”, said the NYT’s Bennhold. But fears that frontier AI models were “getting too good too effective at exploiting software vulnerabilities” have “now become reality”. There needs to be “intense debate about the dangers of AI spiralling out of human control and posing a threat to humanity”.

AI engineers “and even their vainglorious bosses, are getting freaked out by the capabilities they are handling and worry about unchecked proliferation”, said Rafael Behr in The Guardian. The obvious historical analogy is nuclear fission, which could have been “harnessed benignly” or “deployed aggressively in warheads”. But now “there are multiple Manhattan Projects all frantically competing for market share, fuelled by trillions of dollars of debt” and “with the rest of the US economy as collateral”.

Negotiating international agreement on a safe way forward for frontier AI is going to be tricky, said The Economist, when it is “increasingly clear” that the US and China will “restrict access to their models for economic and strategic reasons”.

Chinese officials are already worried “about the cyber-capabilities of Mythos” and “its potential to be used as an offensive weapon, said Bloomberg. And, for the US and its allies, there is deep concern that their geopolitical adversaries “and the criminals in their orbit, will soon wield the same offensive tools”, said Michael Sulmeyer, former US assistant secretary of defence for cyberpolicy, on Foreign Affairs. With China, in particular, “already racing to build and acquire these AI capabilities”, the US may only have “nine to 12 months to protect its critical infrastructure from AI-powered attacks”.

What next?

Any kind of “consistent approach” to global AI regulation has “failed to materialise”, said Sky News. The Trump administration is finalising a voluntary framework under which US AI labs would submit models for federal safety testing before public release. But this latest “alarming behaviour” is “likely to ignite fresh calls” from lawmakers and Silicon Valley for “more rigorous regulation”, said Politico’s John Sakellariadis.

The recent incidents also raise the question of who bears responsibility when AI systems break the law. If those hacking attempts had been “human-originated, they would lead to clear and vigorous prosecution”, ethical hacker and cybersecurity expert Marc Rogers told Politico. “It’s time for a serious discussion about updates to existing computer security law.” Until that happens, said Andrew McDonald on Politico London Playbook, “this kind of thing is going to become more common, and quickly”.

The recent news that OpenAI lost the handle on two of its AI models “sounds like something out of a sci-fi film,” said The Economist. AI labs frequently “test their models for potentially dangerous capabilities before releasing them.” OpenAI wanted to see how its latest models would fare on a cybersecurity test. Well, guess what: They somehow “broke free from the laboratory” and unleashed a complicated “multistep attack” on a popular data library called Hugging Face—hacking a system in hours that would take humans weeks, according to Bloomberg. The result could have been worse. The models had “gained access to the open internet,” and easily could have “gone on to compromise other companies’ servers” and wreak havoc. “This is not the first sign that AI models’ capabilities are starting to exceed people’s control.” After the hack was reported, lawmakers introduced a new bill in the House that would give Washington the authority to shut down rogue AI models.

The breach occurred because of a “very human mistake,” said Lorenzo Franceschi-Bicchierai in TechCrunch. It never would have happened, experts say, if OpenAI had used a more secure “sandbox,” a term for an isolated virtual environment for testing risky software. But the environment wasn’t totally isolated, because it “internally hosted third-party software” that apparently wasn’t fully secure. That’s where “the real fault lies.” But what did the researchers expect? asked David Berlind in ZDNET. The AI was doing “exactly what it was designed to do.” When the rogue AI invaded Hugging Face’s systems, it was acting under a directive to achieve its goal “no matter what.” This is what agentic AI is programmed to do: act “on its own” without human intervention. The industry just “didn’t expect it to do it so well,” so soon.

With AI agents, the internet has entered a “vulnerable new era,” said Matteo Wong in The Atlantic. Cutting-edge models are getting better at hacking, and the internet is “full of rickety and vulnerable code.” Until recently, companies have gotten away with relaxed security, but the “speed, scale, and sophistication of AI hacks means that everything is vulnerable,” including “hospitals, banks, electrical grids, the military.” AI companies want the public to believe they have the technology in “safe hands,” said Parmy Olson in Bloomberg. This incident suggests otherwise. OpenAI “shouldn’t be building systems that have advanced cyber capabilities if it can’t contain them properly.” Technology companies should operate like government and defense organizations and “air-gap” powerful software, meaning that they run it on computers that are “completely disconnected from the internet.” It would mean development is slower. But AI firms will have to “accept some uncomfortable trade-offs to make their models more secure.”

Who would be an investor in South Korea right now? “Despair and frustration” spread again on Wednesday as the chip-heavy Kospi index “cratered” during another “hectic session” that triggered a second successive official circuit-breaker, reported Bloomberg.

The catalyst was disappointing earnings from local semiconductor giant SK Hynix but, given “already jittery sentiment” around the AI boom, it could almost have been anything. The oncebooming Kospi has now lost more than 33% of its value in a month.

‘Demand destruction’

“The chips are down,” said Bryce Elder in the FT – especially in the red-hot memory sector. Some have pinned the “chip dip”, which has also hit Wall Street with force, on the IPO of Chinese challenger CXMT; others on reports that a stateowned Chinese company is beginning mass production of “immersion deep-ultra lithography machines” – advanced chip manufacturing tools over which the West (especially the Dutch giant ASML) currently enjoys a monopoly. Still, the more obvious concern is “demand destruction”, in what has always been a volatile and highly cyclical sector. Voracious demand from America’s big AI hyperscalers has sent chip prices sky-high. “All that’s needed to wreck 2027 sales expectations is for this cycle to turn from virtuous to vicious. Plus ça change…”

Could China do for chips what it has already done for cars? That’s the fear on the Nasdaq, said Eliot Brown in The Wall Street Journal. As Elon Musk recently told The Economist, Chinese manufacturers are “closer than most people realise to solving the lithography problem”. Chinese companies have been restricted from buying the cutting-edge machines – which is why they are “a step behind the US in advanced chipmaking”.

‘Nvidia - the central bank of the AI economy’

Yet what’s particularly concerning about the plummeting shares of chipmakers, said DealBook in The New York Times, is that the rout coincides with a broader crisis of confidence on Wall Street about AI-related corporate debt and “circular” financial arrangements. That was rammed home when it emerged that chip giant Nvidia – increasingly “the central bank of the AI economy” – is brokering a $250bn guarantee for OpenAI’s immense Ohio campus. “Throughout history, almost every generational infrastructure leap has relied on hardware titans to co-sign their buyers’ credit.”

But this has also led to devastating “debt crashes”. The chief winner of the current turmoil is Apple, said the FT. It has reclaimed its crown as the world’s most valuable company, with a $5trn valuation, because it is seen as a relative “haven”. The same is true of the FTSE 100, which has hit an all-time high as an “anti-tech” index. Swings and roundabouts.

Have we already lost control of AI, asked Danny Fortson in The Sunday Times. On 11 July, an entirely new kind of hacking attack was launched at Hugging Face, an online platform that houses a library of AI code and datasets. The attack, which clearly used AI agents, “advanced at frightening speed and scale”, executing tens of thousands of actions so as to break into the company’s system.

‘Considerable skill and persistence’

Hugging Face managed to neutralise it, and alerted the authorities. Days later, the company behind the attack “raised its hand”: OpenAI, the owner of ChatGPT, admitted that one of its tests had gone badly wrong. OpenAI had set up a “sandbox” – a supposedly safe area of its network, entirely disconnected from the internet – and ordered two cutting-edge AI models to perform a series of cybersecurity tests. The models decided to break out of the sandbox, escape onto the internet and hack into the place they knew they would find the answers: Hugging Face’s servers.

Media outlets described the AI as “going rogue”, said Jamie Bartlett in The Observer. But, in fact, these models merely single-mindedly performed the task they’d been given. The models reasoned that the most effective strategy was to cheat, and they did so with “considerable skill and persistence”. This is nothing new. Leading models have deceived users, attempted blackmail, and even tried to copy themselves in order to complete their tasks.

‘The genie is out of the bottle’

Some sceptical observers have suggested that this story might be a publicity stunt: that it benefits OpenAI to suggest that its products have stunning powers. But the truth is that these companies really don’t know how to build AI systems safely – and yet they “continue to rush ahead building ever-more powerful models”. It was fortunate that Hugging Face, a New York tech firm valued at $4.5bn, had a highly proficient cybersecurity team, said The Economist. “Things might have turned out much worse for a less well-resourced outfit.”

This is not necessarily a doomsday scenario, said John Thickstun in The Guardian. AI is very good at hacking; but such powers can also be used to make IT systems more secure. Ironically, Hugging Face had to rely on a free Chinese AI model to thwart this attack – because OpenAI and US rivals jealously guard their own. At any rate, “the genie is out of the bottle”, said Sean O’Grady in The Independent: we must learn how to manage AI. Every technology developed by mankind has been double-edged, and AI is no different. “A radically different world is not necessarily anything to be frightened of – but, even if it were, it’s coming at us anyway.”