Project Glasswing: Anthropic’s AI Hunts for Flaws in Apple Systems

Yesterday, Anthropic unveiled its Glasswing project, a new initiative that employs an advanced model known as Claude Mythos Preview to identify vulnerabilities in operating systems and web browsers. The company claims that this model “exposes a disturbing truth: AI models have now reached a level of programming expertise that allows them to outperform all but the most skilled humans in detecting and exploiting software vulnerabilities.” Anthropic’s system has reportedly discovered thousands of critical vulnerabilities across all major operating systems and web browsers.

Anthropic warns that alternative models to Mythos Preview will soon enable malicious actors to find and exploit weaknesses on a large scale. “The Glasswing project is a critical initiative aimed at harnessing these capabilities for defense purposes.” As such, Mythos Preview will not be released to the public but will instead be available to top-tier partners. Notable collaborators include Apple, AWS, Broadcom, Cisco, Crowdstrike, Google, JPMorgan Chase, The Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks, among others. Additionally, there are 40 more organizations “that design or maintain critical software infrastructures” whose identities have not been disclosed.

Anthropic is committing up to $100 million in usage credits for Mythos Preview as part of this initiative, along with $4 million in direct donations to open-source security organizations. The company also plans to share insights gained from this project to benefit the broader industry.