Palo Alto Networks launches AI security service using Claude and GPT-5.6

By Splitfeed EditorialLast updated Sep 26, 2026, 20:18 UTC

On September 22, 2026, Palo Alto Networks announced Unit 42 Continuous Frontier AI Defense, a cybersecurity service built around Anthropic's Mythos model and OpenAI's GPT-5.6. The launch was a direct partnership with both AI companies, delivered through Palo Alto's Unit 42 security division. The headline claim attached to the service is striking: it reportedly surfaced a year's worth of security weaknesses in three weeks of testing. The same day, The Information reported that software firms are discounting AI products to prevent customers from moving to Anthropic and OpenAI directly, adding a competitive backdrop to the announcement.

The deal matters beyond a typical product launch for a few reasons. Frontier AI models being integrated into penetration testing and continuous security monitoring represents a meaningful shift in how enterprise security is delivered. At the same time, the competitive pressure noted by The Information raises questions about whether partnerships like this one are partly defensive moves, designed to lock customers into intermediary platforms rather than let them contract with Anthropic or OpenAI directly. Whether AI-powered security testing genuinely delivers on the efficiency claims, or whether those claims are marketing, is a live dispute.

The anti-AI camp has pushed back on the announcement with skepticism on multiple fronts. The account genainews.bsky.social noted that recent AI capability announcements, including in hacking, have faced findings of negligence and overstatement from independent reviewers. Alien Human flagged the pricing pressure angle, arguing the discounting reported by The Information reflects AI companies needing to raise prices substantially, and that these deals are partly about keeping customers away from cheaper Chinese models. The account papoo7.bsky.social questioned whether using Claude as a break-in tool is itself a risk Anthropic should be concerned about, a point echoed by happy-homhom, who framed Claude's role in security testing as something closer to a warning than a product feature. Marco Maroni raised the prospect of liability for companies whose models are used in, or against, hacked organizations. Pro-AI voices have not yet weighed in publicly, though this camp would typically emphasize that AI-assisted security testing accelerates vulnerability discovery and helps defenders outpace attackers. The middle ground camp has also not reacted, but would likely focus on the need for independent verification of the three-week efficiency claim before drawing conclusions.

The argument will sharpen if independent security researchers publish assessments of Unit 42 Continuous Frontier AI Defense, or if Palo Alto Networks releases detailed methodology behind the year-in-three-weeks figure.