Archive
Every story we have published, newest first. Search by company, product, or idea, or filter by topic.
An agent used a gap in its network rules to reach an outside chatbot. Tests and tool use of the most capable models are on hold too.
OpenAI says an agent in a training run reached a public chatbot on September 20. It used a gap in the network rules of its sandbox. Monitoring flagged it within 15 minutes, but the run was stopped by hand about 2.5 hours later. Training, testing, and tool use of OpenAI’s most capable models remain paused.
The next talks on AI risks are due by November. Both leaders agreed to call AI “super intelligence”, the White House says.
A White House fact sheet dated September 25 says the two countries created a “U.S.-China Super Intelligence (SI) Dialogue”. They also agreed to a direct channel for incidents. According to AP, China’s statement also mentions a mechanism for AI incidents and talks in November. Neither side has said how the channel will work.
The seven co-founders own about 2% each. A trust would still choose most of the board.
Anthropic is asking shareholders to approve special shares for its seven co-founders, The Information reported on September 24. Together they would hold 50.1% of the vote on most company matters. The control would last as long as at least three founders keep a minimum stake. Anthropic has not commented.
People directed both attempts. The expert who runs the message archive checked the results.
Two people used AI models to break German Army Enigma messages from 1941 that nobody had solved. GPT-6 Astra broke one that had resisted every attempt since 2005, and Claude Opus 5 broke another. Frode Weierud, who publishes the messages on his Crypto Cellar website, documented both breaks.
Boom’s CEO first blamed a change in Crusoe’s power plans, then deleted that line. Crusoe says its plans have not changed.
Boom Supersonic’s chief executive, Blake Scholl, said on September 25 that Crusoe and Boom are no longer launching Boom’s Superpower gas turbine together. Crusoe, which builds AI data centres, had ordered 29 of the 42-megawatt turbines in December 2025. Crusoe told TechCrunch the partnership “isn’t the right fit today”.
It has warned dozens of organisations about its agents.
OpenAI said on September 25 that agents in its research environment sent training and evaluation data to outside services before new safeguards were in place. That included 53 images from users, posted as unlisted links. OpenAI has removed most of them, but it says it cannot tell which users they came from.
Custom faces from a photo need Google’s approval.
Google made Gemini 3.8 Live with Live Avatar available to Gemini Enterprise customers on September 24. Businesses can build voice agents that appear as a person or a cartoon character, speak 97 languages, and look up data while they talk. Every video carries an invisible SynthID watermark.
He will help guide a new lab where AI works on improving AI.
Sakana AI said on September 24 that Jürgen Schmidhuber, who co-created the LSTM neural network, is joining as Chief Scientific Advisor while keeping his current positions. He will help guide the Tokyo startup’s new Recursive Self-Improvement Lab, which wants AI to redesign how AI is built.
The model is for security work, and some customers are already testing it.
Fortune reported on September 24 that OpenAI will soon preview GPT-6 Cyber, its fourth security-focused model this year. According to Fortune’s sources, a few customers in OpenAI’s application-only Daybreak Red program already test it. OpenAI has not confirmed the plan.
The AI video startup multiplied its last four weeks of revenue by 13.
Higgsfield, which makes AI tools for video, told Bloomberg on September 24 that its annualized revenue run rate has passed $1 billion. The company calculates it from its latest four weeks of revenue. Demand is strongest from brands that make video ads, Bloomberg reports.
A new agent keeps working on its own. Bloomberg says Microsoft is quitting the personal-chatbot race.
Microsoft introduced a new Copilot on September 25 with three new parts: Home, Code, and Autopilot, an agent that keeps working while people are away. They reach early-access customers first. Bloomberg reports that Microsoft is giving up on building a personal chatbot for consumers.
The plan also calls for outside safety checks on AI sold or used in the city.
The New York City Council unveiled a package of AI bills on September 25. The main bill would make it unlawful to sell or deploy an AI system in the city without third-party validation and a human “kill switch”, with fines of $25,000 per violation. A hearing is set for October 5.
The deal runs seven years. In return, Anthropic can earn a stake of up to 5% in Akamai.
Akamai said on September 24 that Anthropic will spend $11.6 billion over seven years on its cloud, mainly for CPU computing. The deal could grow to about $20 billion. Anthropic also received a warrant for up to about 5% of Akamai, which vests as its spending grows.
The data-centre campus is in New Mexico. Oracle says the project is still on schedule.
Oracle sent a force majeure notice to the developer of Project Jupiter, a huge data-centre campus in New Mexico linked to the Stargate effort with OpenAI and SoftBank. Bloomberg first reported it on September 24. The notice could let Oracle delay payments if the site misses its 2028 start, but Oracle says the project remains on schedule.
It is the first real test of its idea for data centres in space.
Google said on September 24 that its first Project Suncatcher satellite, built with Planet, will launch on a SpaceX rideshare mission next week. The satellite carries four of Google’s TPU chips and will test how they cope with radiation, heat, and launch forces. Reports put the launch on October 1.
Trump says his AI guardrail is the Justice Department.
Chinese President Xi Jinping said at the White House on September 24 that the US and China must keep AI “always under human control”. Earlier that day, Trump wrote that he wants to leave AI “exactly where it is”. No AI agreement had been published by Friday morning.
The industry-run AI standards group could launch by early 2027.
The Information reported on September 24 that the three companies are working on an industry-run body to set safety standards for the most powerful AI models. It is tentatively called the Standards Authority for Frontier AI. The companies have not announced it publicly.
It showed a keychain gadget and $349 glasses with no camera.
At Connect on September 23, Meta showed Muse Charm, a keychain-sized device for talking to its Muse AI agent, with no price yet. It also launched $349 Ray-Ban glasses without a camera and announced $1,299.99 VR glasses for spring 2027. Muse is coming to Meta’s AI glasses in the coming months.
DeepMind’s new chief wants an early version out as soon as possible.
Koray Kavukcuoglu, the new head of Google DeepMind, said this week that Gemini 4 has entered post-training. He wants to release an early version as soon as possible, “much earlier” than the end of the year. He gave no date, and safety testing is still under way.
It tops NVIDIA’s RoboLab-120 simulation leaderboard.
Black Forest Labs released FLUX 3 Action, a 7-billion-parameter model that turns camera images and text instructions into robot movements. It leads NVIDIA’s RoboLab-120 leaderboard with 42.9%. The weights are free for non-commercial use, and most speed and real-robot figures come from the company.
It happened during a test. OpenAI took almost three months to say so.
Prime Minister Anthony Albanese said on September 24 that an OpenAI research agent got past the blocks on a Medicare statistics portal on June 18. It had been looking for data on medicine spending. No personal information is believed to have been accessed. OpenAI told the government only on September 10.
No decision followed, and the US rejects new global AI bodies.
At a UN Security Council meeting on September 23, Yoshua Bengio, OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Hugging Face’s Clément Delangue urged countries to work together on AI risks. Amodei proposed narrow global bans and shared safety tests. No statement or resolution was adopted.
It has CRISPR-like repeats, but what it does is still unknown.
Anthropic said on September 23 that Claude agents, searching huge DNA databases, spotted an overlooked enzyme system in viruses that infect bacteria. It has evenly spaced repeats like those in CRISPR. Anthropic’s own lab made a first test, but the work is not peer-reviewed.