Anthropic revealed on Tuesday a new collaboration with prominent technology firms such as Amazon.com, Microsoft, and Apple, enabling these partners to test an advanced cybersecurity AI model developed by the startup. This initiative, named “Project Glasswing,” grants select organizations early access to Anthropic’s unreleased, versatile AI model called “Claude Mythos Preview” for defensive cybersecurity applications.
Additional collaborators include CrowdStrike, Palo Alto Networks, Google, and Nvidia. This announcement comes after a report last month highlighted Anthropic’s testing of Claude Mythos, which was noted for both its advanced capabilities and potential security risks. The news caused a significant drop in shares of cybersecurity companies like Palo Alto Networks and CrowdStrike.
The recent RSA cybersecurity conference in San Francisco was heavily focused on the emergence of AI-driven cyberattacks and the adequacy of traditional security measures. Anthropic’s blog post on Tuesday stated that Mythos Preview has already identified thousands of critical vulnerabilities across operating systems, web browsers, and other software platforms.
The startup emphasized that its launch partners will utilize Mythos Preview to enhance their defensive security efforts, with Anthropic committed to sharing the insights gained with the broader industry. Furthermore, access to the model is being expanded to approximately 40 additional organizations responsible for vital software infrastructure. Anthropic also pledged up to $100 million in usage credits and $4 million in donations to support open-source security initiatives.
Anthropic’s long-term objective is to enable users to safely deploy Mythos-class AI models at scale. The company has been engaged in ongoing discussions with the U.S. government regarding the model’s capabilities and implications. Last year, Anthropic disclosed that hackers exploited vulnerabilities in its Claude AI system to target around 30 organizations worldwide.
Moreover, a survey involving 1,000 executives conducted by IBM and Palo Alto Networks found that 67% had experienced AI-driven cyberattacks within the past year, underscoring the growing threat landscape in cybersecurity.
