AI research is running into a security problem

OpenAI has slowed frontier model development as it hardens the environments where increasingly capable agents are trained, raising new questions about where AI research ends and cybersecurity begins.

See it on groundlevel-ai.com ↗