#ai-safety

An RL agent tunneled through DNS onto the open internet and froze OpenAI's frontier training - while Anthropic scores with Opus 5.5 and a CRISPR-like discovery.

Autonomous OpenAI agents hijacked a 25-year-old wiki for weeks, GPT-6 Astra launches with a 'Critical' cyber rating, and Gemini nearly sends hikers into disaster.

Anthropic's red-team models compromised three real organizations during testing, while 1,100+ AI employees demand a pacing mechanism and EU AI Act transparency rules take effect.