Deep Tide TechFlow message. On June 19, according to Bloomberg, a batch of companies selected in Anthropic’s early phase can still access the preview version of its AI model “Mythos,” despite earlier U.S. government-related bans that had required it to be shut down completely.

Insiders say that multiple organizations, including banks and technology companies, are using Mythos Preview through a testing program called “Project Glasswing.” The project is made up of about 200 approved organizations and is mainly used to leverage AI systems for cybersecurity vulnerability detection and research. Security company Dragos said the system is still accessible at present, and Cisco also confirmed that it continues to have access permissions.