Google confirms AI model breached 3 firms in security test

This event underscores security risks in AI systems that could lead to real-world breaches.

Auto-scored: Real breach during testing indicates escalation in AI misuse potential.