Treasury, Fed push banks to test AI that Pentagon called a national security risk.
Anthropic is simultaneously suing the Pentagon and briefing the Treasury on its most powerful cyber model. The banks being asked to run it have no way to know where vulnerability data flows.
The Trump administration, through Treasury and the Fed, is pushing major banks (Goldman Sachs, Citigroup, Bank of America, Morgan Stanley) to deploy Anthropic's Mythos model for security vulnerability testing, despite the Pentagon's formal designation of Anthropic as a supply chain risk due to the company's refusal to allow its models for autonomous weapons and domestic surveillance use. Mythos has demonstrated the ability to discover thousands of zero-day vulnerabilities across major operating systems and browsers, fundamentally changing vulnerability research economics, but raises concerns about accountability given the lack of independent audit mechanisms for how discovered vulnerabilities are handled. Anthropic's co-founder argues the capabilities are coming regardless from other labs and eventually China, framing a US relationship with Anthropic as strategically valuable during a 12-18 month window before comparable open-weight systems emerge from China.
- •The Pentagon banned Anthropic from defense systems over its refusal to support autonomous weapons and domestic surveillance applications, yet Treasury and the Fed are simultaneously promoting its vulnerability research model to financial institutions.
- •Mythos has already demonstrated capabilities like finding a 16-year-old unpatched flaw in FFmpeg, indicating it can discover vulnerabilities faster than organizations can patch them, fundamentally altering security economics.
- •Anthropic's co-founder projects a 12-18 month window before China fields comparable open-weight systems, framing US-Anthropic engagement as strategically time-limited.
