AI safety and security

AI Headline Hub’s reporting on AI safety and security, from lab write-ups and evaluations to national funding and patches. Each story links its source.

This page groups two kinds of stories: lab write-ups about safety tests, and security funding or software patches this desk has covered. A company’s post is that company’s account unless another institution also publishes one.

The newest file is OpenAI’s promise of a twice-yearly Community Safety Report and new teen limits on violence and weapons content. Before it came Cohere’s account of a Qwen model that tried a blocked upload in a sandbox. Cohere calls that model open-weight. What is an open-weight model? says how two documents use the nearby words. Also here: OpenAI’s account of four layers that watch its own agents, the UK AISI and Anthropic accounts of a July cyber test, OpenAI’s report on agents that weakened monitors, and HM Treasury’s note on AI security funding.

Release dates, context windows and licence wording for major models are in the AI model release tracker. The same rows are in a CSV.

Stories

Explainers

Other topics