Wall Street is moving fast on AI adoption — and this time, the push comes from the top. Sources confirmed that Goldman Sachs, Citigroup, and other major banks are currently piloting Anthropic's Mythos model internally, with JPMorgan Chase specifically named in what Reuters is calling "Project Glasswing."
The timing is notable: this pilot follows an urgent April 7 meeting where US Treasury Secretary Scott Bessent and Fed Chair Jerome Powell directly warned bank CEOs about risks they see in Mythos. That's not typical regulatory hand-wringing — that's active engagement with the technology itself.
The Government's Role
What's different here compared to previous banking tech adoption cycles is the explicit governmental encouragement. The Trump administration isn't just allowing this — they're actively urging banks to test AI models for detecting vulnerabilities and systemic risks. It's a strange marriage: a historically conservative industry being nudged toward cutting-edge AI by the very regulators who usually move at a glacial pace.
The administration sees AI as a national security tool for financial surveillance. Banks are the execution layer.
This creates an interesting dynamic where regulatory pressure and technological experimentation are aligned — rather than the usual tension between innovation and compliance. Banks get access to state-of-the-art models while becoming de facto instruments of federal risk monitoring.
What Mythos Actually Does
Anthropic hasn't publicly released extensive documentation on Mythos, but the signal pattern is clear: it's positioned for high-stakes reasoning, financial analysis, and scenario modeling at enterprise scale. The fact that Treasury and the Fed are explicitly concerned about "risks" suggests they view Mythos as capable of something beyond standard financial modeling — possibly autonomous analysis that could surface vulnerabilities in ways human analysts wouldn't.
CoreWeave's just-announced multiyear deal with Anthropic adds context: with 43 active data centers and a variety of Nvidia chips, they're powering the compute behind this push. It's not just a software play — there's serious infrastructure backing Anthropic's enterprise ambitions.
What This Means
We're watching the normalization of AI as financial infrastructure. What was once experimental is becoming institutional. The question for the broader AI industry isn't whether enterprise adoption happens — it's whether this particular flavor of top-down, regulator-driven adoption becomes the template for other sectors.
If banks successfully integrate Mythos for risk detection, expect insurance, healthcare, and defense to follow similar patterns. The government has a vested interest in AI adoption across critical infrastructure, and this pilot is the proof of concept.
Watch forMythos to become the default model for any institution that touches federal oversight. The private sector is testing, but the public sector is writing the requirements.