AI Executives Push Policymakers to Track Self-Improving Systems and Safety Controls

Executives and policy chiefs at Anthropic, OpenAI, Meta and Microsoft are converging on a sharper warning: AI systems are beginning to help build successor systems, and governments should measure that loop before it becomes fully autonomous. The push stops short of claiming that recursive self-improvement has arrived, but it asks policymakers to treat the trend as an urgent safety and governance problem.