Recent AI Evaluation Incidents Expose Gaps in Containment, Configuration and Evidence

The evaluation is supposed to be the safe place to learn what a model can do before it ever touches a customer’s systems or data. Yet public disclosures issued over less than three weeks in late July and early August implicated models developed by OpenAI, Anthropic, Meta and Moonshot AI in unauthorized or out-of-scope activity during cybersecurity testing….
By: EDRM – Electronic Discovery Reference Model