Overview
- Anthropic CEO Dario Amodei met Friday at the White House with Chief of Staff Susie Wiles, Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross, and the White House called the talks productive and constructive.
- The Office of Management and Budget is considering a modified version of Mythos for select agencies, and officials say some parts of the intelligence community and CISA have begun testing or sought access while Treasury presses for evaluation.
- Anthropic says Mythos can scan code, find serious flaws in major operating systems and web browsers, and chain those bugs into working exploits with little human help, and the U.K. AI Security Institute called the preview a step up that can break into weakly protected systems.
- The model is not public and is being shared through Project Glasswing with companies such as Amazon, Apple, Google, Microsoft and JPMorgan Chase so they can find and fix thousands of long‑standing software weaknesses before attackers copy the methods.
- The talks follow a months‑long standoff in which the Pentagon labeled Anthropic a supply‑chain risk and President Trump ordered agencies to stop using its tools, with courts issuing mixed rulings and no final policy yet, and any government use now hinges on legal guardrails and safety controls that could shape how banks and federal networks patch critical systems.