Breaking

Palo Alto Networks launches AI-driven service to find and fix cyber exposures

September 22, 2026 at 08:28 EDT

Security
AI Agents
Sales & Business

Palo Alto Networks is launching a subscription service that uses multiple cyber-capable AI models to continuously probe customer environments for exploitable weaknesses and recommend fixes. The September 22 rollout puts Anthropic's Claude Mythos 5, OpenAI's GPT-5.6-Cyber and open-weight models into a Unit 42 workflow designed to find attack paths before intruders do.

September 22, 2026 | Palo Alto Networks

AI models move into continuous cyber defense

The new service uses multiple cyber-capable models to find exploitable weaknesses, validate attack paths and push prioritized fixes before attackers can chain the same flaws.

40%

No single model topped this vulnerability catch rate

10%

Upper bound on Mythos 5 and GPT-5.6-Cyber overlap

14,090

Confirmed vulnerabilities in NOVA research

95.0%

GPT-5.6-Cyber advanced-request completion rate

Same test, different completion rates

95.0%

GPT-5.6-Cyber

2.0%

Daybreak Blue

1.5%

GPT-5.6 Sol

Each orange segment represents roughly five percentage points; the small columns show how sharply the specialized cyber model changes the response rate in advanced defensive tasks.

How the service works

Scope

Agree on testing coverage

Discover

Map exposures with AI

Validate

Verify attack paths

Remediate

Send fixes to workflows

Defensive upside

Multi-model routing can widen coverage and help teams focus on validated attack paths rather than isolated alert volume.

Guardrail load

Cyber-capable agents still need tight scope, controlled access and expert review before findings become operational changes.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.