Palo Alto Networks (NASDAQ: PANW) announced Unit 42 Continuous Frontier AI Defense, an agentic offensive security service that uses AI models from Anthropic and OpenAI to continuously identify and remediate enterprise security exposures.

The service integrates Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6-Cyber models through a proprietary multi-model harness that routes security tasks to the most suitable AI model. It is designed to provide continuous vulnerability testing, attack path validation across web applications, APIs, cloud infrastructure, source code repositories, and network assets, along with prioritized remediation guidance.

According to the company, the service was developed over six months of internal testing and more than 100 customer engagements, backed by a $17 million investment in research and development. During internal deployment, Palo Alto Networks said the system identified a year's worth of exposures within three weeks. In customer assessments, exposures were found in 100% of cases, with 37% rated high or critical severity. The company also noted that more than two in three exposures found in third-party applications had no known CVE designation.

This service expands on Unit 42's Frontier AI Defense, which launched in April and offered point-in-time exposure analysis. In August, Palo Alto Networks announced it would add OpenAI's GPT-5.6-Cyber and Anthropic's Claude Mythos 5 to that offering.

Sam Rubin, Senior Vice President of Unit 42, said: "AI has created an asymmetric advantage for threat actors against organizations trying to defend at human speed. Modern cybersecurity requires machine-speed defense."

The service is available worldwide on an annual subscription basis, with pricing options that vary depending on which AI models are selected. All subscription tiers include access to the multi-model harness architecture.