Palo Alto Networks: "Mithos discovered vulnerabilities in three weeks equal to a year's worth of manual penetration testing"

Palo Alto Networks says tests of frontier AI models show AI-assisted attacks can match long manual testing in finding vulnerabilities and move faster to attack chains.

The company unveiled its Frontier AI Defense initiative to counter risks from AI-driven vulnerability discovery, exploit chaining, and automated intrusion.

The effort brings together partners including OpenAI, Anthropic, Accenture, Deloitte, IBM, NTT Data, PwC, and Unit 42.

The findings suggest AI is moving beyond a hacker aid and could sharply accelerate real-world offensive operations, raising the bar for enterprise defense.