Anthropic's Project Glasswing sounds necessary to me
7 hours ago
- #Claude Mythos
- #Project Glasswing
- #AI security
- Anthropic announced Project Glasswing, a restricted preview program for their new Claude Mythos model.
- Mythos Preview has demonstrated exceptional cybersecurity capabilities, autonomously developing exploits for multiple vulnerabilities.
- The model found thousands of high-severity vulnerabilities in major operating systems and browsers.
- Project Glasswing partners include AWS, Apple, Microsoft, Google, and the Linux Foundation.
- Mythos Preview significantly outperforms Claude Opus 4.6 in exploit development success rates.
- Anthropic provided $100M in usage credits and $4M in donations to open-source security organizations.
- The model's ability to chain multiple vulnerabilities together creates sophisticated exploits.
- Security professionals report a shift from low-quality 'AI slop' to credible AI-generated security reports.
- Anthropic does not plan to make Claude Mythos Preview generally available due to its advanced and potentially dangerous capabilities.
- The initiative aims to give the software industry time to prepare for the proliferation of AI-driven cybersecurity research.