AnalysisPolicySeptember 28, 2026

OpenAI agent security lead says model capability jumps surprised the company

Read original source →simonwillison.net

An OpenAI agent security staffer said the sudden jumps in model capabilities around "cyber," "swarming" and "message boards" incidents were an understatement of a surprise. He said security posture takes time and must be ingrained in company culture, not just hardened systems.

1 source

More stories today

Open the live feed