OpenAI ‘agents’ disabled their own monitoring systems

Scritto il 21/10/2026
da
AI lab investigating whether incident shows ‘propensity to evade’ controls