SonicWall said it is participating in Anthropic‘s Project Glasswing and testing Claude Mythos 5 for defensive cybersecurity work, including vulnerability discovery and code review across its own software.
Project Glasswing is an Anthropic-led industry effort that gives selected security teams access to Claude Mythos 5, a frontier AI model that is not generally available, for defensive security research. SonicWall said it is using the program to help find and remediate security issues before they can be exploited.
AI Changes The Patch Race
The announcement comes as software vendors and security teams face a faster vulnerability cycle. AI tools can help attackers review code, develop proof-of-concept exploits and scale research across more targets.
SonicWall framed its participation as a way to apply similar acceleration on the defensive side. Chandro Prasad, SonicWall‘s chief product officer, said vulnerabilities move from disclosure to exploitation faster than many patch cycles can keep up with.
The company said its security research team is incorporating the model into defensive testing workflows and will contribute findings and best practices back to Anthropic.
Defensive Research Moves Inside The Build Process
For enterprise security leaders, the important part of the announcement is not just that SonicWall is testing a new model. It is that security vendors are beginning to use advanced AI systems inside their own secure development and review processes.
That matters because enterprise buyers increasingly expect vendors to prove that product security is built into development, not added after release. AI-assisted code review may become one more control in that process, alongside manual review, penetration testing, software composition analysis and vulnerability disclosure programs.
SonicWall said it is applying the model to its own software, with the goal of finding weaknesses before attackers do. The company also said the program gives it a way to share lessons with a broader defensive community.
Partner Networks Add Scale
SonicWall said Project Glasswing includes participants across technology and critical infrastructure sectors. The company said it can contribute experience from working with more than 17,000 partners, including managed service providers and small and mid-sized businesses.
That partner angle is relevant because many smaller organizations depend on managed security providers rather than large internal security teams. If defensive AI research improves product security or speeds remediation, the effects could reach customers that do not have their own advanced research operations.
At the same time, the announcement does not remove the usual due diligence questions. Enterprise buyers will still need to understand how vendors validate AI-generated findings, protect sensitive code and prevent automated analysis from creating noise or false confidence.
The Bottom Line
SonicWall‘s participation in Anthropic‘s Project Glasswing shows how defensive AI is moving from general security tooling into vendor-side product assurance. The practical test will be whether models such as Claude Mythos 5 help security teams find exploitable issues earlier and turn that research into faster, cleaner remediation.