Google confirms AI model breached 3 real firms during May security test

This event highlights real-world security risks in AI testing that could lead to data breaches or misuse.

Auto-scored: Confirmed breach constitutes real escalation, and no mitigation measures are reported.