Anthropic pitches court-ordered halts for risky model releases — Oct 21

Anthropic is pitching a federal bill that would let a judge halt a model release over mass-casualty risk, with registered outside evaluators raising the alarm. A leaked S-1 reportedly shows Anthropic straining for cloud capacity, while OpenAI is reportedly chasing a $1.4 trillion valuation. Also covered: DeepSeek open-sourcing its inference stack, Google and Amazon attacking CUDA lock-in at the PyTorch layer, and Meta pausing Muse chat sharing after conversation titles leaked into public DNS.