Wall Street Banks Test Anthropic's Mythos AI
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Goldman Sachs, Citigroup, and other banks are testing Anthropic's Mythos model internally after Trump administration officials encouraged them to use it to detect vulnerabilities, per Bloomberg.
- JPMorgan Chase is the only bank specifically named in Project Glasswing, the initiative that prompted the broader bank testing.
- Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened an emergency meeting of banking CEOs to address the cyber risk, an unusual joint Treasury-Fed intervention on a single AI model.
- The UK's AI Security Institute (AISI) tested Mythos and called it the most capable cyber model it has ever evaluated, and the first to complete its hardest cyber-range end-to-end.
- The Bank of England raised alarm in parallel, with the Telegraph reporting UK officials had described the AI as "too dangerous to release."
- Canadian bank executives and regulators also met to discuss the risks raised by the model, per the Globe and Mail.
- Senator Richard Blumenthal (D-CT) cited the situation to push his bipartisan legislative framework for AI guardrails, arguing the dangers are growing more visible by the day.
Why it matters: Bessent and Powell's joint CEO summons is an unusually coordinated Treasury-Fed move on a single AI vendor, elevating Mythos from a procurement question to a systemic financial-sector concern. The Bank of England and UK AISI sounding parallel alarms signals transatlantic regulatory pressure on Anthropic and on any bank that deploys the model.
