Artificial intelligence company Anthropic has demonstrated new capabilities for AI-assisted cybersecurity after its Claude Opus 4.6 model identified 22 vulnerabilities in Mozilla’s Firefox browser, including 14 classified as high severity, during a two-week security collaboration with Mozilla.
Anthropic’s engineering team began analyzing Firefox’s JavaScript engine before expanding their review across additional areas of the browser’s codebase. The vulnerabilities were discovered through AI-assisted analysis of the open-source code, and most have already been addressed in Firefox 148, released earlier this year. While the model proved highly effective at identifying security issues, it was far less successful at generating exploit code, with the team producing only two proof-of-concept exploits after using approximately $4,000 in API credits.
The announcement comes as Claude remains widely available to enterprise customers despite escalating tensions between Anthropic and the U.S. Department of Defense, which recently designated the company a supply-chain riskafter Anthropic declined to grant unrestricted military access to its AI systems.
Major technology platforms including Microsoft, Google, and Amazon Web Services confirmed that Anthropic’s models will continue to be available through their cloud platforms for non-defense workloads. Microsoft said Claudewill remain accessible through products such as Microsoft 365, GitHub, and Azure AI Foundry, while Google Cloudconfirmed similar availability.
Dario Amodei, CEO of Anthropic, said the designation applies narrowly to Pentagon contracts and does not prevent commercial customers or contractors from using Claude in non-defense applications.
