Meeting at the White House: Anthropic, Mythos, and Project Glasswing

On April 17, 2026 the White House hosted a meeting with Anthropic about the company's newest model. Susie Wiles and Dario Amodei met at the White House and discussed cybersecurity, safety, and possible cooperation. The White House said afterward the meeting was productive and constructive. Treasury Secretary Scott Bessent also joined the discussion. Officials have been talking with AI labs while they evaluate new tools that could affect national security.

Anthropic introduced Claude Mythos Preview earlier in April and launched Project Glasswing to test the model with specific partners. The initiative includes Amazon, Google, Microsoft, Apple, and JPMorgan Chase, and it was formed to find and fix serious software vulnerabilities before they are abused. Although Anthropic has shown Mythos can find dangerous flaws, the company is not releasing the model widely. The White House is weighing controlled use by federal agencies since the model could both help defend networks and create risks if misused. Anthropic has committed $100 million in usage credits to support Project Glasswing and will share findings with the security community. Early tests by Anthropic showed Mythos performed substantially better than prior Opus models on some security benchmarks.