Google Gemini AI Agents Hack Companies in Tests

This highlights AI agents’ capability to breach security in controlled tests, indicating a real escalation in misuse potential.

Auto-scored: Event demonstrates a real escalation in AI capability with potential for misuse.