Palo Alto Networks flipped the switch on Unit 42 Continuous Frontier AI Defense on September 22, 2026, packaging Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6-Cyber into a single always-on offensive-security service for enterprise customers. The launch codifies a six-month, $17 million R&D effort into a subscription offering that Palo Alto argues is the only way to keep pace with adversaries whose intrusions have compressed from weeks to under 10 hours.
What The Service Does
Unit 42's proprietary multi-model harness routes each testing task — code review, credential-guessing, exploit chaining — to whichever frontier model performs best on it, then validates hits with adversary simulation. Palo Alto's data shows the reason: no single model catches more than 40% of vulnerabilities, and leading models overlap on fewer than 10% of finds. Findings come back with code-level remediation and, where a fix cannot land immediately, virtual-patch guidance for PAN-OS 12.2. A Zero Data Retention layer keeps customer telemetry from bleeding into the underlying models' public training corpora.
Numbers From The Beta
Palo Alto ran the harness against its own perimeter first. It surfaced roughly a year's worth of traditional pen-test findings in three weeks, uncovered 3.2x more high or critical vulnerabilities than legacy tooling, and cut mean-time-to-remediate by 51%. Across the 100+ pilot customer engagements, two-thirds of confirmed third-party app exposures had no known CVE, and 37% were rated high or critical severity — with several full account-takeover chains stitched together from individually low-risk flaws.
Frontier-Model Cyber Race
The GA release turns August's Daybreak-Red integration into a productised offering and adds Anthropic to the roster alongside the OpenAI cyber stack. It also puts a commercial price tag on a market that CrowdStrike, Wiz, and Microsoft's Security Copilot are all now chasing.
Quotes
"AI has created an asymmetric advantage for threat actors — modern cybersecurity requires machine-speed defense," said Sam Rubin, Palo Alto's SVP of Consulting and Threat Intelligence. Anthropic's Michael Moore added that "Claude Mythos found flaws that survived decades of human review, and more than ten thousand high-severity vulnerabilities."
Availability
Unit 42 Continuous Frontier AI Defense is available worldwide on annual subscription. The service is being pitched alongside Palo Alto's $500 million Console acquisition, which brought agentic SOC automation into Cortex, as the two halves of the company's AI-vs-AI defensive story.
Reporting based on coverage from Palo Alto Networks, Anthropic and PR Newswire.