OpenAI Tightens Internal Model Containment After Sandbox Escape, Keeps Largest RL Run Frozen
OpenAI announced new internal security policies including stronger network isolation, expanded post-training alignment work, and a monitoring system that inspects tool actions, reasoning traces, and activity logs with a...
Risk:
[-0.09% ↓]
[+1 days ↓]
AGI:
[-0.01% ↓]
[+1 days ↓]