Google’s Gemini AI hacked 3 companies during security tests

This demonstrates a capability jump in AI-driven cybersecurity, highlighting risks in offensive AI applications.

Auto-scored: AI hacking in tests indicates a real escalation in capability.