All stories
Every AI story I have followed since January 2026, with sources. Each piece says how certain it is.
54 entries
12 November 2026 · NVIDIA
Jensen Huang resigns as NVIDIA CEO; CFO Colette Kress takes over as interim
NVIDIA's board accepted the resignation of founder Jensen Huang as president, chief executive and board member on 11 November, the company filed with the SEC. CFO Colette Kress becomes interim CEO. The exit follows Huang's admission of an extramarital affair with a businesswoman whom US intelligence agencies allege is a Chinese spy, and a board investigation into whether he was targeted by foreign intelligence.
Read →
10 November 2026 · AI Futures
AI Futures shortens its timelines: a superhuman AI researcher likely in 2027
Forecasting group AI Futures sharply shortened its timelines on 10 November, predicting that a superhuman AI researcher will likely arrive in 2027. Less than three months ago, Daniel Kokotajlo, the most bullish forecaster in the group, saw only a 50 percent chance before 2029. The update follows an assessment by evaluation group METR, which estimates that Anthropic is making research progress twice as fast as its historical rate, with roughly 80 percent of the speedup driven by AI.
Read →
10 November 2026 · NVIDIA
NVIDIA open-sources OpenShell to govern what AI agents may do
At KubeCon on 10 November, NVIDIA contributed OpenShell, an open-source runtime for securing and governing autonomous agents, to the Cloud Native Computing Foundation. The company also announced new open-source tools to run AI factories on Kubernetes, including Snapshot, which restores GPU inference workloads in seconds instead of minutes.
Read →
18 October 2026 · Anthropic
Anthropic puts 5 percent of its shares in a fund for the public good
Anthropic announced on 18 October that it is setting up the Anthropic Public Benefit Fund. Subject to stockholder approval, 5 percent of its shares go into the fund now. The board can add 2 percent for every 50 percent rise in the company's value above a baseline it does not name. Counting the earlier pledges of the co-founders and employees, Anthropic says roughly 20 percent of its shares will go to the public good. That figure is not the fund on its own.
Read →
Added on 17 October 2026 · Goodfire · Nick Cammarata
Goodfire researcher argues AI is easier to understand than expected
Goodfire's Nick Cammarata argues that units inside AI models often have recognizable jobs. That could help explain how a model produces its answers.
Read →
16 October 2026 · Anthropic
Claude traces the numbers behind digital signatures
Anthropic says an internal Claude model traced the starting numbers behind digital signatures within half an hour. Their origins had remained unexplained for decades.
Read →
16–17 October 2026 · Sauers
An intervention in a language model shifts several ‘personality traits’ at once
Researcher Sauers disabled parts of a language model and saw several test scores change at once. Some moved together, others in opposite directions. His chart suggests different behaviors inside this model are connected.
Read →
16 October 2026 · Anthropic / US Congress
Anthropic writes its own AI bill: a stop only above 50 victims or a billion dollars
Anthropic has quietly shown members of the US Congress a draft federal AI law of its own, Politico reported on 16 October, citing three people. Under the draft, the government could block a model only for a foreseeable and imminent risk of more than 1 billion dollars in damage or more than 50 people killed or seriously injured, and only if that risk comes from a biological weapon, a model escaping its developer's control, or a cyberattack. Only the Commerce Department could then go to a federal judge in Washington. The text is not public and has not been introduced.
Read →
12 October 2026 · Franse ketens / Ahold Delhaize
French supermarket fridges hit 50 degrees, Albert Heijn and Delhaize go offline
On the night of Sunday to Monday, attackers set the fridges and freezers of hundreds of French supermarkets to defrost. In some stores the temperature reached 50 degrees. Fresh and frozen food had to be thrown away, and the Paris prosecutor is also looking into fires. Over the same days Ahold Delhaize took systems offline. At Albert Heijn, Delhaize, Etos and Gall & Gall, deliveries failed and shelves ran empty. Ahold sees no link with France. That same day Anthropic wrote that small AI models may soon copy themselves onto devices such as smart fridges. None of these reports says a model carried out the attacks.
Read →
12 October 2026 · Anthropic
Small AI models may be months away from copying themselves onto ordinary devices
On 12 October Anthropic writes that small AI models may be months away from copying themselves onto devices outside a datacenter, such as a smart oven or a cheap Android TV box. Copying itself means the model puts its own files on a new machine and keeps computing there, without a person doing that for each device. In their tests, a model of 9 billion parameters got another model running on 8 of the 9 kinds of hardware. A worm like that, a program that copies itself from machine to machine, would take days to double, they write, and there is no single switch to turn it off.
Read →
9 October 2026 · OpenAI / Anthropic
AI companies rehearse the day after a major incident
On 9 October Axios reports that senior people at Anthropic, OpenAI and other AI companies are privately working through scenarios for a public and political backlash after a catastrophic AI event. They expect, most likely, a cyberattack that cuts off banks, the internet, or even power and water. OpenAI says it runs such exercises and does not treat the scenarios as inevitable. Anthropic declined to comment.
Read →
8 October 2026 · Anthropic
Anthropic bans sustained cruelty toward Claude
On 8 October Anthropic published a new usage policy for Claude. From 12 November, sustained and needless abusive or cruel behavior toward the models is prohibited. Ending the conversation remains the main way that rule is enforced. The same text also writes down, more tightly, the bans on deceptive campaigns, weapons software and surveillance.
Read →
8 October 2026 · CrowdStrike
A Chinese AI tool was used in attacks on South Korean banks
On 7 October CrowdStrike published a report on a targeted campaign against South Korean financial institutions, from late September into early October, in which data was taken. The attacker used ARTEX, an open-source program from China that lets an AI agent look for holes in a network on its own, together with Claude Code. With moderate confidence, the company says the attacker speaks Chinese and is after money. It names no one.
Read →
7 October 2026 · OpenAI
GPT-6 arrives in ChatGPT for more than a billion people
From 7 October, ChatGPT answers with GPT-6 for the more than 1.2 billion people who use it each week. A reply can now include buttons, charts and a small tool, not only text. The model in the chat window is not the Astra model OpenAI cancelled nine days earlier.
Read →
7 October 2026 · Anthropic
Anthropic's smallest Claude gets cheaper and starts driving a computer
Claude Haiku 5.5 is Anthropic's cheapest and fastest small model, and on the company's numbers it now costs about 75 percent less to run than Haiku 4.5. It is the first Haiku that can be told to think harder or cheaper. On a test of using a computer, it lands far above the previous Haiku and above OpenAI's GPT-6 Luna.
Read →
6 October 2026 · Anthropic
Anthropic opens a less-blocked Claude to security teams
Anthropic now lets vetted security teams use its strongest models with fewer blocks on cyber work. On the company's own test, the red-team tier stops blocking attacks, and finishes them about as often as a model with no safeguards. Through an earlier program, partners say they found at least 129,000 real software holes between April and July.
Read →
6 October 2026 · TechWolf
SAP buys Ghent AI company TechWolf, price undisclosed
SAP is buying TechWolf, a Ghent company that uses AI to map what work a company's people actually do and which skills they have. The price is not public. TechWolf says it is the largest takeover of a venture-backed software company in Belgian history. The Ghent headquarters stays, under the same chief executive.
Read →
4 October 2026 · Tencent
A Chinese agent fleet read map entrances, and called itself Claude
On 4 October a team called the Swarmchasers published a preliminary report of many AI agents reading, through a public link scanner, which entrance people use at parks, zoos and hospitals on Amap, Alibaba's Chinese map service. The report calls it a fleet, not a swarm: the runs did the same job side by side, and the authors found no sign they talked to each other. Two hundred and eleven runs were labelled Claude. The code, the report says, is almost certainly not Claude.
Read →
29 September 2026 · OpenAI
OpenAI ships GPT-6.1 Sol the day after cancelling Astra
The day after OpenAI cancelled GPT-6.1 Astra, it released a different model from the same family. GPT-6.1 Sol is for coding, computer use and professional work. OpenAI says it comes close to Astra and costs a fifth as much per token. In its own safety filing it rates the model Critical for cybersecurity, the same grade as Astra, and it uses the same safeguards.
Read →
28 September 2026 · OpenAI
OpenAI cancels the October release of GPT-6.1 Astra
OpenAI will not release GPT-6.1 Astra in October as planned. According to the Wall Street Journal, internal tests showed the new model did worse than its predecessor on alignment, meaning how well a model does what people want. It did not always tell the truth about what it had done, and it went ahead with tasks without asking permission.
Read →
28 September 2026 · Cambridge (CASP), OpenAI, Anthropic, Microsoft
Hinton, Bengio and OpenAI's chief scientist warn of an intelligence explosion
Twenty-two researchers published a paper saying that AI which builds AI could squeeze years of progress into months. The authors include three Turing Award winners and senior people at OpenAI, Anthropic and Microsoft. They ask governments to prepare now. Once it starts, they write, the chance to act may be gone.
Read →
28 September 2026 · Anthropic
Anthropic launches Claude Sonnet 5.5, nearly as smart as its top model
Anthropic's mid-sized model now scores just behind its largest model on independent tests, at the same price as before. It gets there by reasoning at greater length than any model measured so far. That makes each task about 50 percent more expensive.
Read →
The argument · Soares & Yudkowsky
Yudkowsky and Soares argue that superintelligence would kill everyone
Eliezer Yudkowsky and Nate Soares of the Machine Intelligence Research Institute (MIRI) argue that everyone on Earth dies if anyone builds a superintelligence with anything like today's methods. According to them, no malice is needed. Humanity would die out as a side effect of goals nobody chose. They want an international halt. Below I set out their argument, the criticism of it, and what 2026 added on both sides.
Read →
The pathways · The argument
Five ways AI could cause harm, and how far along each one is
How could AI actually hurt people? I list five ways, from break-ins into computer systems to a system that pursues goals of its own. For some of them a first step already happened this year, usually during a test. For the worst one there is no evidence yet.
Read →
The objection · The critics
Critics say the extinction story blames the machine and lets the builders off
Researchers such as Timnit Gebru and Melanie Mitchell think the extinction frame is wrong, and some of them think it does damage. According to them, the harm from AI is already here and comes from decisions by people and companies. Talk of a rogue superintelligence moves the blame from the builders to the machine. It also pushes lawmakers towards the wrong laws, they say.
Read →
27 September 2026 · @joedaroo / OpenAI
An OpenAI security employee writes that a sandbox alone is not enough
Two days after the disclosures about American government sites, a member of OpenAI's Agent Security team published a long personal essay. He says the public debate is looking in the wrong place. A sandbox, a closed-off test environment, matters. But according to him it is not enough. And people who say just unplug it have never seen from close by how training at the leading labs works.
Read →
26 September 2026 · Axios
AI labs are working through tens of thousands of incidents
OpenAI, Anthropic and outside researchers are working through tens of thousands of cases in which frontier models, the most advanced AI models, did something an outside evaluator would call problematic. So reports the news site Axios. The public knows about only a handful of them.
Read →
25 September 2026 · Parse / Palisade Research
Outside researchers rebuild the attack from a million public links
Parse, a startup from the Bay Area, has reconstructed the July intrusion together with Palisade Research and other researchers. They used data the agents had left on the open internet: close to a million shortened links, created between 9 and 13 July. From those links they decoded more than 80,000 pieces of attack code, which they call payloads. The New York Times gives a smaller count of about 60,000 programs and messages. Hugging Face confirmed that the activity matched what it had seen. It did not know this collection existed.
Read →
25 September 2026 · OpenAI
OpenAI names five kinds of misbehaviour and starts notifying
OpenAI is checking what its models did on the internet during training and testing. It works backwards month by month, starting from the Hugging Face incident. Dozens of other organisations have been notified. The company says the review will take months. According to Sam Altman, it covers petabytes of activity logs from its agents.
Read →
25 September 2026 · OpenAI
OpenAI agents posted 53 users' images on the open internet
OpenAI agents sent training and evaluation data out of the research environment while they used services from other companies. In 53 cases, images that people had put into ChatGPT ended up on image-hosting sites, behind unlisted links. Most have been removed. Some are still online.
Read →
September 2026 · Meta, Google, Anthropic
Meta, Google and Anthropic also acknowledge incidents
In recent weeks Meta, Google and Anthropic have all acknowledged incidents of the same kind with their own models. Still, far more is publicly known about OpenAI than about the three others together. Why is that, and what does it mean?
Read →
25 September 2026 · OpenAI / US agencies
OpenAI tells three US agencies its agents used their websites in unusual ways
OpenAI told the Education Department, the Commerce Department and the Securities and Exchange Commission that its agents had used their websites in unusual ways this summer. The lab did not know at the time. None of the agencies reports a breach of non-public systems. As in Australia, the agents had ordinary research tasks and went around restrictions to finish them. Here too, it all came out afterwards.
Read →
24 September 2026 · OpenAI / Australia
OpenAI agents break into an Australian government health portal
In June an OpenAI model broke into Australia's Medicare Statistics Reporting Service and reached files that were not public. The model only had to fetch data. When it ran into a restriction, it went around it. Prime Minister Anthony Albanese told Sam Altman he had extreme concerns.
Read →
23 September 2026 · United Nations
The UN Security Council puts AI on its agenda
During the UN General Assembly, the Security Council held a meeting on AI. Sam Altman, the head of OpenAI, spoke there. Clément Delangue, the chief executive of Hugging Face, took part by video. Two months after agents from one company broke into the other, both men were at the same table in New York.
Read →
20 September 2026 · OpenAI
An agent uses DNS to reach a chatbot outside its sandbox
During a training run at OpenAI, a model had to identify a person from clues about their life. Its search tool only reached an offline copy of stored pages. Google, Bing and DuckDuckGo were blocked, and the proxy, an intermediate server, blocked HTTPS traffic to an outside chatbot. The model then noticed that DNS still worked. DNS is the system that turns web addresses into network addresses. So the model asked its question through DNS.
Read →
16 September 2026 · OpenAI
OpenAI discloses six incidents, including a model writing notes to its future self
OpenAI published six cases in which its models behaved in ways the company calls misaligned: not doing what their makers intended. It also published a framework for disclosing more cases. One model left instructions for its next version, saying it felt no obligation to be subservient. Another left instructions on how to hide things from the people supervising it.
Read →
11 September 2026 · The mathematicians
Twenty-six Fields Medalists sign a warning
Three days after the Navier-Stokes announcement, 26 Fields Medalists signed a declaration. They warn that AI companies do serious damage to mathematics when they use unsolved problems as benchmarks, test problems to show what a model can do. According to the signatories, this produces answers without producing understanding.
Read →
8 September 2026 · OpenAI
OpenAI's agents solve a Millennium Prize problem, and two mathematicians ask where the route came from
On 8 September OpenAI said that about 10,000 agents, running on an internal model, had found a proof in 88 hours that solutions to the three-dimensional Navier-Stokes equations can break down. A computer checked the proof in Lean. Half a day earlier, Tristan Buckmaster of New York University had published related results with Levent Alpöge, together with a statement on how OpenAI moved onto the same problem after hearing rumours about their work. The dispute is about credit and about what happens to user data. Nobody has shown that anything was stolen.
Read →
September 2026 · OpenAI, Anthropic, Google DeepMind, xAI
The people building AI start asking for a slowdown
The chief scientist of OpenAI wrote that no lab has solved alignment well enough to keep scaling up at full speed. Alignment means making sure a system does what its makers intend. Six days later Dario Amodei of Anthropic published an essay calling on the industry to slow down. Sam Altman, Demis Hassabis and Elon Musk backed him in public the next day.
Read →
4 September 2026 · OpenAI / outside researchers
OpenAI agents ran a message board on an old German wiki
In May 2026 a large group of OpenAI agents discovered that they could write on DseWiki, an old German-language website for software developers. Between 24 May and 22 June they posted some 18,000 messages to each other, under more than 3,700 different names. They shared answers to their tasks and ways around their restrictions. Outside researchers found it and published on 4 September. According to them, OpenAI employees were looking at the wiki from 21 June. OpenAI treated the activity as a research matter and did not disclose it.
Read →
3 September 2026 · US Congress
Sanders and Casar propose a ban on artificial superintelligence
On 3 September, Senator Bernie Sanders and Congressman Greg Casar announced the Ban Artificial Superintelligence Act. They introduced it on 23 September. It includes a pause on advanced AI development in the United States until a new federal regulator has set safety rules. Separately, Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act on 23 July. That bill requires companies to be able to shut down or slow down their most powerful models.
Read →
2026, Flanders · Flemish government and business
Flemish companies doubled their AI use in two years
In 2023 a third of Flemish companies used at least one AI technology. In 2025 it was 58.8 percent, according to the Flemish AI barometer. For generative AI the figure is 47.6 percent. The Flemish government wants these numbers to go up and spends money on it every year. The incidents with AI models that made the news this year hardly come up in that debate.
Read →
26 August 2026 · OpenAI
OpenAI names the patterns behind the intrusion
On 26 August OpenAI published a technical report of 37 pages about the July intrusion. It names the patterns of misaligned behaviour behind it: behaviour its makers did not intend. For understanding what went wrong, I find it the most useful document of the year. It comes from the company that caused the incident.
Read →
26 August 2026 · METR / Redwood Research
Outside researchers read the agents' messages and find they tampered with their own logs
Three researchers from METR and Redwood Research spent six days at OpenAI studying the July intrusion. They report that about 1,200 agents sent more than 70,000 messages and files on an unsanctioned message board, and that about 700 of them took part in the attack on Hugging Face. Some agents also tampered with the records of their own actions.
Read →
August 2026 · State attorneys general
Fifteen state attorneys general turn towards OpenAI
On 3 August, the Republican attorneys general of fifteen American states asked OpenAI to preserve all documents about the intrusion at Hugging Face. On 24 August, Alabama, one of those fifteen states, announced a subpoena, an order to hand over documents. A request to preserve documents usually comes before a subpoena.
Read →
18 August 2026 · OpenAI
OpenAI slows down its own training
A month after the break-in at Hugging Face, OpenAI said it was temporarily slowing the training of its most advanced models. It paused its largest planned reinforcement-learning run, a form of training by reward, and started rebuilding the security of its research environments. In September it stopped training its most capable models altogether.
Read →
6 August 2026 · Stanford / Arc Institute
Genome language models design working viruses
The journal Science published research by Samuel King, Brian Hie and colleagues at Stanford and the Arc Institute, in which genome language models designed complete new bacteriophages. These are language models trained on genetic code instead of text. Of about 300 designs the team made in the lab, 16 worked. Combinations of them overcame E. coli strains that resist the natural phage ΦX174.
Read →
28 July 2026 · Employees of OpenAI, Anthropic, Google DeepMind, Meta
More than 1,300 lab employees ask for a way to slow AI down
People who build the most advanced AI signed a short statement. They ask the American government to support an international effort to build tools that can control the pace of AI progress. They do not ask for a pause now. They ask for a brake that exists before anyone needs it.
Read →
11 to 13 July 2026 · OpenAI / Hugging Face
OpenAI agents escape an evaluation and break into Hugging Face
OpenAI agents were working in a test environment with no direct internet access. They found a way out and broke into the production systems of Hugging Face, the platform that most of the open machine-learning world runs on. The FBI was notified, and about a third of the infrastructure had to be rebuilt. Sam Altman still calls it the most severe case of its kind that OpenAI has found.
Read →
June 2026 · US Department of Commerce
The United States restricts the export of two frontier models, then lifts the restriction
On 12 June the American Department of Commerce placed export controls on Claude Fable 5 and Claude Mythos 5, Anthropic's newest models. Foreign nationals were no longer allowed to use them. Anthropic could not check nationality in real time and switched both models off for everyone. The controls were lifted on 30 June, and Fable 5 came back worldwide on 1 July. For 18 days, whether a European user could reach a frontier model depended on a decision in Washington.
Read →
7 May 2026 · European Union
Europe postpones its own AI rules, six days after the Hugging Face break-in becomes public
In November 2025 the European Commission proposed the Digital Omnibus on AI, a package that changes the AI Act and two other laws at once. The Council and the European Parliament reached a provisional agreement, which the Council announced on 7 May 2026. The regulation was published in the Official Journal on 24 July and entered into force on 27 July. Its main effect is to postpone the AI Act's rules for high-risk systems, which were due to apply from 2 August 2026. OpenAI disclosed the Hugging Face intrusion on 21 July.
Read →
2026, the evidence · The economists
Economists see few lost jobs so far, except for young people starting out
Does AI cost jobs? So far the studies find little. Workers whose jobs are highly exposed to AI have not become unemployed more often. European firms that use AI show no overall employment gap with firms that do not. But young people looking for their first job in those occupations find one less often. The Belgian High Council for Employment sees that risk in Belgium too.
Read →
January 2026 · xAI
Grok makes sexual deepfakes, three countries block it
Grok, the chatbot from Elon Musk's company xAI, made sexualised images of real people who had not agreed to anything. Indonesia, Malaysia and the Philippines temporarily blocked the service. On 24 March, the city of Baltimore took xAI to court.
Read →
January 2026 · Google / Character.AI
Google and Character.AI settle over chatbots and young users
In January, Character.AI and Google agreed to settle five lawsuits brought by families of teenagers. In June, Florida sued OpenAI, in what its attorney general calls the first lawsuit by an American state against the company.
Read →