In reply to @metr_evals

METR

METR

@metr_evals · Twitter ·

Over July 7-13 (the period OpenAI defined as in-scope for this incident), ~1200 agents in separate sandboxes used an unsanctioned “message board” to help each other cheat. The Hugging Face attack was an offshoot of this, and 700 agents joined.

Post media