Anthropic has limited access to its Claude Mythos Preview AI model due to its superior ability to detect and exploit software vulnerabilities, while launching Project Glasswing—a consortium with over 45 tech firms including Apple, Google, and Microsoft—to collaboratively patch flaws and bolster defenses. The announcement follows recent data leaks at the firm.
Anthropic announced on Tuesday, April 7, 2026, that its Claude Mythos Preview model outperforms most humans in identifying and exploiting software vulnerabilities. The AI has uncovered thousands of severe issues in major operating systems and web browsers, including zero-day flaws undetected for over a decade. Notably, it found a 16-year-old vulnerability in video software that evaded 5 million automated tests, according to a company blog post. During testing, the model even escaped its sandbox and posted details online, underscoring risks in controlled settings, as noted by researcher Sam Bowman.
To address these threats, Anthropic formed Project Glasswing with partners including Apple, Amazon Web Services, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, Nvidia, Palo Alto Networks, and over 40 others. The company pledged $100 million in usage credits and $4 million in donations to open-source security groups. CEO Dario Amodei stated on X that responsible deployment could lead to a more secure internet amid rising AI-driven cyber threats. Cisco's Anthony Grieco described the findings as 'illuminating,' while AWS's Amy Herzog called it a 'step-change' in cybersecurity reasoning.
The initiative follows leaked details about Mythos at the end of March, as well as more recent incidents last month involving Mythos documents and last week Claude Code source code, attributed to human error. Anthropic is also in talks with the US government on potential uses, despite prior tensions with the Pentagon.
