Treasury warns banks on Anthropic AI Claude Mythos risk
Get the Geopolitics newsletter
Daily geopolitics — wars, elections, sanctions, the diplomatic moves that move markets. Free.
- Treasury Secretary Scott Bessent convened a meeting with CEOs of Bank of America, Citi, and Wells Fargo in Washington, DC, warning that deploying Anthropic’s Claude Mythos Preview could expose sensitive customer data to cyberattackers.
- Federal Reserve Chairman Jerome Powell attended the briefing, underscoring the Fed’s concern about cyber threats to the financial system.
- Anthropic’s Claude Mythos Preview is a new AI model designed to uncover software security vulnerabilities beyond human detection, and Anthropic has limited its release to a coalition of 40 companies called Project Glasswing.
- Project Glasswing includes JPMorgan Chase, which plans to use the model to evaluate next‑generation AI tools for defensive cybersecurity across critical infrastructure.
- JPMorgan CEO Jamie Dimon was invited to the briefing but did not attend because of prior travel plans.
- Kevin Hassett, director of the National Economic Council, told Fox News that officials are taking every step to ensure safety, including Anthropic holding back a public release until the risks are resolved.
- The Trump administration is in a legal dispute with Anthropic after the Department of Defense designated the company a “supply chain risk” for limiting AI use in war.
Why it matters: The warning forces banks to halt or tightly control deployment of Claude Mythos Preview, preserving customer data security but delaying AI‑enhanced vulnerability detection; regulators gain leverage over AI risk management, while Anthropic’s restricted rollout curtails its commercial rollout and fuels a legal clash with the Trump administration over supply‑chain designations.
