What happened
TechCrunch reported on April 12, citing Bloomberg, that Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell brought bank executives into a meeting this week and encouraged them to test Anthropic’s Mythos model for vulnerability detection. JPMorgan Chase had already been identified as an initial partner, and Bloomberg said Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley were also testing the model.
That is notable because Anthropic only just announced Mythos and said access would be limited in part because the model appears unusually strong at finding security vulnerabilities. TechCrunch also highlighted the policy contradiction: Anthropic is still fighting the Trump administration in court over a Department of Defense decision to label the company a supply-chain risk.
Why this matters
This is not just another enterprise pilot. It is a sign that frontier AI policy in Washington is becoming increasingly fragmented and tactical. Instead of one coherent government view of a lab, different arms of the state are treating the same company in different ways depending on the immediate institutional need.
For Anthropic, that matters because regulated-sector adoption can build legitimacy even while national-security disputes remain unresolved. For banks, it matters because the model is being positioned not as a generic chatbot, but as a specialized system for security-relevant work inside high-stakes institutions.
The strategic read
The deeper significance is that AI competition is moving away from a purely general-purpose assistant race and toward workflow wedges that are easier to justify economically. If a model can help banks identify vulnerabilities, the sales story becomes concrete very quickly: lower risk, better internal tooling, faster remediation.
But the policy picture gets stranger as a result. A government that cannot decide whether a lab is too risky for one supply chain may still decide that the same lab is useful enough for another critical domain. That kind of selective adoption may become normal as agencies optimize for local objectives rather than a unified AI doctrine.
Bottom line
The Mythos story is bigger than one bank test. It suggests government-backed AI adoption is entering a messier phase in which political distrust, commercial urgency, and sector-specific usefulness can all coexist at the same time.
Source note
Source: TechCrunch, "Trump officials may be encouraging banks to test Anthropic’s Mythos model," published April 12, 2026.