The Schizophrenia of the Trump Administration on Anthropic: A National Security Threat for the Pentagon, the Treasury Pushes It Towards Banks
The United States Treasury and the Federal Reserve are pushing major American banks to test Claude Mythos, the new model from Anthropic for cybersecurity. Meanwhile, Anthropic itself is still at war with the Department of Defense, which two months ago labeled it a risk to the national supply chain. The same company is thus blacklisted by the Pentagon, but at the same time recommended by Scott Bessent and Jerome Powell in one of the clearest photographs of the many contradictions of the AI policy of the Trump administration.
Project Glasswing: The Consortium
As previously reported, Project Glasswing brings together about 50 organizations to which Anthropic has granted early access to Claude Mythos Preview, including Amazon Web Services, Apple, Google, Microsoft, Nvidia, Cisco, CrowdStrike, Palo Alto Networks, and JPMorgan Chase. In addition, Anthropic has allocated up to $100 million in usage credits for partners and $4 million in donations to open-source security organizations. What matters here is who is pushing the banks to use it, and why right now.
The Pentagon Paradox
The story between Anthropic and the DoD has roots in a $200 million contract signed in July 2025, for which, last January, the Pentagon requested the removal of two contractual restrictions: the ban on use in fully autonomous weapon systems and the ban on deployment for the mass surveillance of American citizens. CEO Dario Amodei refused: "We cannot in good conscience concede to the Pentagon's requests," he publicly stated. The administration's response was swift and severe: on February 27, 2026, Trump ordered all federal agencies to cease using Anthropic's products. Secretary of Defense Pete Hegseth designated Anthropic as a "supply chain risk" for national security, a classification usually reserved for foreign entities suspected of ties to adversarial governments. Internal DoD documents obtained during litigation reveal that the designation was motivated by the company's "hostile demeanor through the press": not a technical security issue, but a punishment for how it had behaved publicly. A Pentagon official accused Amodei of having a "God complex." Trump, on Truth Social, called Anthropic a "radical left, woke company."
Anthropic responded by opening two simultaneous legal fronts. On March 26, 2026, Judge Rita Lin issued a preliminary injunction blocking the designation, writing in a 43-page decision that "nothing in the appropriate statute supports the Orwellian idea that an American company can be branded as a potential adversary and saboteur of the USA simply for dissenting from the government." However, the D.C. Circuit rejected the request to suspend the designation during the appeal: the result is that Anthropic is excluded from DoD contracts but can continue to operate with all other federal agencies.
Banks and International Regulators
With the scenario outlined, Treasury Secretary Scott Bessent and Fed Chairman Jerome Powell convened this week the heads of JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley, urging them to use Mythos to identify vulnerabilities in their systems. A recommendation that does not come from an industry blog, but from the top figures of American financial oversight. This level of institutional pressure leaves banks with little room for choice, regardless of the Pentagon's position towards the supplier.
The wave has also extended across the Atlantic. The Financial Times reported that officials from the Bank of England, the Financial Conduct Authority, and HM Treasury are in discussions with the National Cyber Security Centre to assess the potential vulnerabilities highlighted by Mythos. Representatives from major banks, insurers, and British exchanges will be updated and informed within two weeks. If Mythos has indeed found thousands of zero-days in the most widely used operating systems and browsers, any financial infrastructure running on those same software stacks is potentially exposed.
Anthropic co-founder Jack Clark, speaking at the Semafor World Economy in Washington, tried to separate the two issues: "We have a specific contract issue, but I want to emphasize that we are deeply committed to national security. The government needs to be informed about these developments. We are already discussing Mythos with them and will also talk about future models." A posture that acknowledges the conflict with the DoD without renouncing dialogue with the rest of the administration.
The Structural Contradiction
The Mythos affair exposes a coherence problem that goes beyond mere institutional discontent. The Trump administration has classified Anthropic as a national security threat because the company refused to remove the restrictions on the use of AI in autonomous weapons and mass surveillance, which the Pentagon considered an operational obstacle. In the Treasury and Fed's reading, apparently, they are not. The result is that the left hand of the executive blacklists what the right hand points to test on the country's critical financial system.
However, the situation exposes Anthropic to an advantage: every bank that adopts Mythos deepens the company's integration into the national critical infrastructure, making the supply chain risk designation increasingly difficult to sustain in court and on a political level. The breadth of the Project Glasswing consortium, with Microsoft, Google, Apple, and Amazon publicly aligned with Anthropic even in legal proceedings, builds a network of dependencies that the Pentagon can hardly isolate without systemic consequences.