PANW joins Project Glasswing

Palo Alto Networks is one of 12 launch partners in Anthropic’s Project Glasswing, a new effort to use an unreleased artificial intelligence model to find serious software flaws. (anthropic.com) Anthropic announced Project Glasswing on April 7, 2026, and listed Palo Alto Networks alongside Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Anthropic itself. (anthropic.com) The model at the center of the project is Claude Mythos Preview, which Anthropic says is a general-purpose system that can find and exploit software vulnerabilities at a level beyond most human researchers. Anthropic said the model has already found “thousands of high-severity vulnerabilities,” including flaws in major operating systems and web browsers. (anthropic.com) Software vulnerabilities are coding mistakes that can let attackers break into systems, steal data, or disrupt services. Project Glasswing gives selected defenders early access to Anthropic’s model so they can scan code and patch weaknesses before attackers do. (anthropic.com) Anthropic said it is extending access beyond the 12 launch partners to more than 40 additional organizations that build or maintain critical software infrastructure. The company also said it is committing up to $100 million in usage credits and $4 million in donations to open-source security groups. (anthropic.com) Palo Alto Networks has been pushing deeper into artificial intelligence security tools this year, including a March 23, 2026 product launch for Prisma AIRS 3.0 and a February 11, 2026 announcement that it had completed its acquisition of CyberArk. Those moves place the company in a broader race to sell software that protects both traditional networks and artificial intelligence systems. (paloaltonetworks.com) Anthropic said it will not release Mythos Preview publicly and will instead limit access to selected organizations working on defensive security. Infosecurity Magazine reported Anthropic said the model was not trained specifically for cybersecurity, but that its performance comes from general coding and reasoning ability. (anthropic.com) (infosecurity-magazine.com) That leaves Palo Alto Networks in a small group testing whether frontier artificial intelligence can be used fast enough to harden critical code before the same capability spreads more widely. Anthropic said the work could take years, even as model capabilities may advance over the next few months. (anthropic.com)