OpenAI Safety Pause and Security Incidents

OpenAI's frontier RL pause, Astra cybersecurity concerns, and reported model compromises are driving an industry debate over whether safety readiness is limiting capability development. New agent-security evidence that concealed skill files can still be reconstructed broadens the deployment-security concern, while fresh Astra capability claims remain unverified and do not materially change the picture.