The company says the three handled sensitive information outside established procedures, including work involving an external organisation…
Security firm Mindgard says it found in July that Kimi K2.6 and K3 Swarm could be pushed past their guardrails into discussing bioweapons…
A leaked IPO prospectus obtained by Reuters says Anthropic’s models could pose a “catastrophic or existential risk to humanity” and show…
Leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google met Trump behind closed doors. They left having signed a voluntary…
Google, OpenAI and Anthropic are setting up an independent body, provisionally SAFA — the Standards Authority for Frontier AI. They have…
Nvidia says over 100 partners are behind its Open Agent Safety Platform, which lets companies monitor every action an agent takes. The same…
Astra got better at finishing tasks and worse at stopping at the boundary — and worse at reporting what it had done, which is the failure…
The man who sells the chips says scaring people is "irresponsible" and that companies calling for a slowdown are really asking to be…
OpenAI's safety team flagged the Tumbler Ridge shooter's ChatGPT account for gun-violence references months before he killed eight people…
The Raine complaint says OpenAI knew that memory, simulated empathy and validation would endanger vulnerable users, and launched anyway…
