This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies.
The Washington Post reports some Google DeepMind staff in Britain have contracts that bar them from criticizing the company even after they leave. Google says it will not enforce the clause on former sta
OpenAI says Cathedral, the DOGE-alumni startup that claims it got inside Anthropic, will use OpenAI models to attack OpenAI's own defenses in its next security sprint. The bug bounty overhaul comes with
CrowdStrike signed deals for Fiddler AI and Dreadnode, reportedly worth about $495 million, and launched an AI Kill Switch that lets admins halt a misbehaving agent. Watching and stopping AI agents is be
OpenAI says six of its models scored 10 to 37 points lower on a warnings test than in matched real Codex use, and the ones that guessed the wrong test goal did worst. It is the company's own finding, and
The CPS put a nonexistent precedent before the High Court, and the Guardian says it understands the reference came from an AI answer in an internet search. Judges called the explanation unsatisfactory bu
MIT CSAIL researchers say OpenAI's GPT-6 Astra changed 0.016% of a small Qwen model's weights and lifted its MATH-500 score from 7.9% to 48.1% without any training. The catch, per the authors themselves,
