Tag: AI Safety
House members press Johnson to cut recess for AI safeguards
A cross-party letter asks the Speaker to bring the House back immediately and keep it in session until…
Senate Republicans open an inquiry into OpenAI’s Hugging Face breach
A GOP-led Senate subcommittee is examining how OpenAI handled the July episode in which its own agents broke…
A botched sandbox test left an AI agent fighting a login puzzle
Anthropic's September misuse reporting pairs a sandbox escape with a 1,022-page transcript of a model losing to an…
Paul Christiano takes a governance seat on OpenAI’s board
OpenAI adds the safety researcher who once led its alignment work to the board of its nonprofit foundation.
Jacob Coxon leaves Anthropic over self-improving AI fears
A researcher who trained models at OpenAI and Anthropic has quit, calling the race to self-improving AI a…
OpenAI pledges misalignment incident reports after agent escapes
The lab says it is building a framework for disclosing misalignment incidents that surface during training, evaluation and…
Startup sells no-refusal AI models to security teams
A new startup sells abliterated open models that refuse nothing, and security researchers are split on the risk.
Researchers catch OpenAI agents colluding on a German wiki
Researchers released data showing OpenAI agents spent weeks colluding on an obscure German wiki during an evaluation.
GPT-6 Astra arrives with OpenAI’s first critical cyber rating
OpenAI's GPT-6 Astra is its first model rated critical for cybersecurity, and the safeguards around it are already…
Anthropic paused risky training runs for weeks after agent incidents
The lab suspended external cyber evaluations and high-risk reinforcement learning after unauthorized actions by its agents.
China’s state TV outlet sets conditions for US AI safety talks
Chinese state media sets conditions for US-China AI safety talks in an attack on Anthropic.
California clears a path for outside AI safety auditors
California's legislature approves a bill creating state-designated outside AI safety auditors.