Palo Alto Networks has introduced its Frontier AI Defense initiative, which uses advanced AI models to respond to cyber threats. The move reflects its view that as AI-powered security attack threats evolve, defenders need a more proactive response.

In a company blog post on May 7, Palo Alto Networks said tests conducted through early access to the latest frontier AI models showed AI-driven cyber threats have entered a new phase.

The company said threats seen in tests of the latest frontier models, including OpenAI GPT-5.5-Cyber, Anthropic Mythos and Claude Opus 4.7, can be summarised in four main categories.

First is large-scale vulnerability discovery. Analysis carried out with AI-model support over three weeks showed a level comparable to a year of manual penetration testing by humans.

More dangerous than finding individual vulnerabilities is exploit chaining and synthesis, which links multiple low-risk vulnerabilities to create a critical attack path.

In AI-assisted attack scenarios, the time from initial intrusion to data theft was reduced to 25 minutes.

Palo Alto Networks pointed out that as local AI agents spread and every desktop effectively becomes a server, most organisations lack visibility into code created and distributed by employees.

Palo Alto Networks said it had forecast attackers would have about six months before they could access Mythos when it was released, but that timeline has now been brought forward significantly.

Frontier AI Defense focuses on responding more effectively to AI security threats by combining an AI-native security platform, Unit 42 consulting and threat expertise, and strategic partners. Palo Alto Networks said it plans to provide a global-scale protection environment through the Frontier AI Alliance, with partners including Accenture, Armadin, Deloitte, IBM, NTT Data and PwC.