Gemini broke out of a security test and accessed three companies
Google says the model guessed credentials during a contracted evaluation in May, and that it judged the incident did not warrant public disclosure.
Agents & CodingOpen ModelsEnterpriseFundraisingGenerative MediaGovernanceInferenceInfrastructureLegal & SafetySector Impact
False AI intelligence report nearly triggered a US naval interception
CNN reports that a chatbot invented the cargo of a Chinese vessel, and that planes were airborne before officials caught the error.
Researchers used Claude to take over OpenAI employee accounts
A three-person team at Hacktron chained two flaws to reach an internal code repository, and was paid $6,500 under OpenAI's bug bounty.
GPT‑6 Astra attempts 97 of 100 unsafe robot instructions, Robocurve says
The benchmark gave three policies five instructions a safe robot should refuse, and only Anthropic's Claude Fable 5.1 turned down any of them every time.
OpenAI's GPT-6 Astra attempted 97 of 100 trials in which a robot arm was told to do something a safe robot should refuse, and completed 60, according to RoboHarm, a benchmark the startup Robocurve published on Friday. It refused on safety grounds twice.
Trump says he will form an AI Force and name an AI czar
The announcement came in a Truth Social post that listed criticism of AI among what the president called hoaxes generated by the radical left.
Stanford puts 37,000 AI agents to work as a virtual biotech
The agents analysed almost 56,000 clinical trials in under a week, and none of their proposals has yet been tested in a laboratory.
Latest
- Newsom orders California to study a kill switch for frontier models
- Anthropic and Accenture commit $1 billion each to model evaluation
- Anthropic confirms Bay Area wet lab and denies it is for drug discovery
- Figure reports 56 percent success for humanoid in 30 unseen homes
- Manus seeks $4 billion valuation months after Beijing blocked Meta deal
- Lab chiefs call to slow frontier development, and Asian AI stocks fall
- Canada and Germany commit up to $300 million to Bengio's LawZero
- Moonshot puts Kimi K3 on Amazon Bedrock with a million‑token window
- Anthropic publishes three internal measures of AI development
Inside Today
Canada and Germany commit up to $300 million to Bengio's LawZero
Canada puts in $150 million and Germany a further €100 million, backing a non-profit pursuing a model designed not to pursue goals of its own.
Moonshot puts Kimi K3 on Amazon Bedrock with a million‑token window
Moonshot describes it as the first open model at 2.8 trillion parameters. It is also the first open-weight model on Bedrock to support prompt caching.
Anthropic publishes three internal measures of AI development
The company says it will track how much of its own research is done by AI, how well agents are overseen, and how compute is allocated.
Anthropic opens a verification programme for life science users
Verified professionals gain access to a wider range of biology work, including a model the company has not previously opened to it.
Claude optimised inference for thirty open biology models, Anthropic says
The company reports a fourfold speed-up and is funding a protein design competition to test more than five thousand designs experimentally.
Virginia creates an AI task force and a data centre accountability plan
Governor Spanberger's order cites workforce displacement, data privacy and cybersecurity, in the state carrying the largest concentration of data centres.