Palo Alto Networks Turns to Claude and GPT Models for New AI Defense Service

Palo Alto Networks will launch a cybersecurity service built on AI models from Anthropic and OpenAI, the company said on Tuesday, in a direct answer to attackers who now use the same technology to hunt for weaknesses in corporate networks.

The service, called Unit 42 Continuous Frontier AI Defense, relies on cyber-focused AI models including Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6-Cyber, alongside open-weight models. Palo Alto Networks did not disclose the subscription pricing in its announcement.

The launch responds to a shift in the threat environment. Hackers are increasingly using AI to find and exploit weaknesses in corporate networks, forcing cybersecurity firms to build defenses that can detect and respond to threats more quickly. Palo Alto Networks is betting that the same class of models can be turned around and put on defense.

The new offering is designed to continuously test web applications, application programming interfaces and cloud infrastructure. The goal is to help customers identify vulnerabilities and possible attack paths as their digital environments change — a moving target that static, periodic security assessments have struggled to cover.

Detection is only half the product. Palo Alto Networks said the service would also provide guidance on fixing security gaps, including code-level fixes and virtual patching options. That positions the offering as a remediation tool rather than a pure scanning product, a distinction that matters for security teams with limited engineering bandwidth.

The commercial model ties cost directly to model choice. The service will be available globally through annual subscriptions, with pricing based on the mix of OpenAI, Anthropic and open-source models selected by customers, the company said. That structure effectively lets customers trade off performance and price across frontier and open-weight models — an unusual pricing approach that reflects how expensive frontier-model inference remains.

For Palo Alto Networks, the launch extends its Unit 42 brand — best known for threat research and incident response — into a continuous, software-driven security product. It also deepens the company's reliance on third-party foundation models from Anthropic and OpenAI rather than models built entirely in-house, a bet that specialized cyber-tuned versions of general-purpose AI will outperform what any single vendor can train alone.

The move signals where the cybersecurity market is heading: as attackers weaponize AI, vendors are racing to industrialize AI-driven defense. Expect rivals such as CrowdStrike, Microsoft and Zscaler to answer with comparable offerings, and expect the mix of frontier and open-weight models under the hood to become a purchasing decision in its own right.