Agent Safety and Governance Acceleration

EU AI transparency rules, OpenAI agent escape, Patronus AI $50M funding, DeepMind unsafe states, and U.S. Marine Corps agent registry signal rapid policy and safety developments. New tools like UNMASK for spurious correlation mitigation and claim-level reliability assessment improve robustness. OpenAI President urges enterprise deployment with 10-step security playbook.