The Trump administration has achieved something genuinely remarkable: it is simultaneously trying to destroy Anthropic in court and begging Wall Street to use Anthropic's technology. Not in some vague, bureaucratic way where different agencies have different opinions. Treasury Secretary Scott Bessent and Fed Chair Jerome Powell personally summoned the CEOs of JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley this week to urge them to test Anthropic's new Mythos model. The same Anthropic that Defense Secretary Pete Hegseth designated a "supply chain risk" and tried to ban from every federal contract.

Goldman Sachs CEO David Solomon told reporters he is "hyper-aware" of Mythos's capabilities and is working "closely" with Anthropic on cybersecurity. Three countries, the US, Canada, and Britain, have now held emergency banking meetings over the model. And today, Anthropic co-founder Jack Clark went public at the Semafor World Economy event in Washington, confirming the company is in active talks with the Trump administration about Mythos and future models. "We have a narrow contracting dispute, but I don't want that to get in the way of the fact that we care deeply about national security," Clark said.

Let's be clear about the timeline here. In February, Hegseth gave Anthropic CEO Dario Amodei a Friday deadline to drop its safety restrictions on autonomous weapons and mass surveillance. Amodei refused. Hegseth designated the company a supply chain risk. Trump ordered federal agencies to stop using its technology. A Pentagon official accused Amodei of having a "God complex." Two months of legal warfare followed.

Then Anthropic released Mythos, a frontier model so capable at finding cybersecurity vulnerabilities that the company chose not to release it publicly. Through Project Glasswing, roughly 50 organisations got access, including Apple, Google, Microsoft, Nvidia, CrowdStrike, and JPMorgan. The model found thousands of zero-day flaws across every major operating system and browser. Anthropic committed $100 million in usage credits and $4 million in donations to open-source security projects.

Suddenly, the same administration that branded Anthropic a national security threat started telling the country's systemically important banks to adopt its product. The cognitive dissonance is spectacular. The Pentagon says Anthropic is too dangerous to work with. The Treasury says Anthropic is too important not to work with. Both positions exist in the same government, in the same month, about the same company and the same technology.

Here is what is actually happening. The Pentagon dispute was never about national security. It was about Anthropic refusing to let its AI be used for autonomous kill chains and warrantless surveillance. The Mythos recommendation is about actual national security: protecting the financial system from cyberattacks that could cascade across the global economy. The Pentagon wanted compliance. The Treasury wants capability. And Anthropic, which refused to bend for the military, is now being courted by the regulators.

This matters beyond Anthropic. It sets a precedent. A company can refuse the military's demands on ethical grounds, get blacklisted for it, and still end up as the government's preferred solution for the exact kind of security the Pentagon claimed it was protecting. The "supply chain risk" label looks increasingly absurd when the Treasury and Fed are actively pushing that same supply chain into the banking sector.

Clark's public comments today are Anthropic calling the bluff. They are telling the administration: we will cooperate on national security, we will share what Mythos finds, but we will not remove guardrails on weapons. And the administration, it turns out, needs them more than it wants to admit. The Pentagon can keep its court case. The money just showed up at a different door.