tech
April 10, 2026
Anthropic’s Mythos Preview AI model too powerful for public release and partners with big tech
Anthropic says its new Mythos Preview AI model is too powerful for public use, launching Project Glasswing with major tech partners and consulting the US government

TL;DR
- Anthropic's new AI model, Claude Mythos Preview, is deemed too powerful for public release by company executives.
- The model can act autonomously, reason like a security researcher, and generate exploits for detected vulnerabilities.
- Anthropic is discussing the model's capabilities with the US government.
- The company is launching Project Glasswing, a partnership with over 40 organizations including Apple, Amazon, Microsoft, Google, and NVIDIA, to test the model's ability to find and fix software vulnerabilities.
- Internal materials and secret code for Anthropic's AI have been accidentally leaked previously.
Continue reading the original article