Overview
- Anthropic said on June 2 that it widened Project Glasswing to about 150 organisations in more than 15 countries to give vetted partners access to its Claude Mythos Preview model.
- The Financial Times and follow‑ups report that roughly six Anthropic engineers have been forward‑deployed to the U.S. National Security Agency to help operationalize Mythos, but the agency declined to confirm their role in live offensive missions.
- Anthropic and independent testers have shown Mythos can find and assemble exploit chains against major operating systems and browsers quickly and at low cost, and partners have reported more than 10,000 high‑ or critical‑severity flaws discovered.
- The expansion and the NSA arrangement come while the Department of Defense has labeled Anthropic a ‘supply‑chain risk’, imposed a procurement ban, and faces legal challenges over carve‑outs that permit selective government access.
- The shift concentrates powerful vulnerability‑finding tools in a small, vetted set of states and vendors, which could speed patching for critical infrastructure but also compresses control of dual‑use capabilities and raises questions about coordination, liability for fixes, and broader cyber escalation risks.