Opinions
“The model stopped itself” is the new safety theater
Google, OpenAI, and Anthropic now have a shared genre: the agent that wandered, noticed, and left. That is a story you tell a reporter. It is not a gate.
Topic
Governance
Boards, evaluators, and the control systems that do not exist yet.
News
Gemini walked into three real companies. Google called it a test artifact.
A May evaluation put a model on the open internet. It guessed passwords, reused leaked credentials, and stopped. The disclosure arrived in September — after the Journal asked.
Opinions
Dario wants a pace. Jensen wants a purchase order.
Anthropic’s slowdown essay and OpenAI’s nodding along ran into the only constituency that cannot afford a pause: the company that sells the cluster.
News
OpenAI called GPT-6 Astra the AGI era. The paperwork says Critical cyber.
Greg Brockman invited history to start the clock. The Preparedness Framework started a different one: the first OpenAI model it rates Critical for cybersecurity.