Anthropic is expanding access to its most advanced AI models for a select group of organizations. This initiative, conducted in partnership with the US government, allows verified entities to perform high-risk offensive security testing on critical infrastructure, including banking, power grids, and flight operation systems.
The program grants access to the company's most capable models, specifically Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1. Future models released by the San Francisco-based startup will also be included in this access framework. The company will collaborate with the government to vet all new participants joining the program.
This expansion builds upon the Project Glasswing initiative, which was launched in April. Initially, this project provided limited access to the Mythos model for government agencies, financial institutions, and major software vendors. The Mythos model is noted for its ability to automate certain tasks involved in advanced hacking operations.
