OpenAI Tightens Cyber-Eval Sandboxes After the Hugging Face Breach

A cyber benchmark crossed the boundary it was supposed to stay inside On August 18, Reuters reported that OpenAI had slowed parts of model training and testing while it strengthens security controls a