Bloomberg reported that earlier this week, U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell showed a significant shift in tone during a meeting with bank executives: they began encouraging banks to use Anthropic's newly released Mythos model to detect security vulnerabilities in the financial system.
From "Warning of Risks" to "Empowering Tools"
Previously, regulatory officials had expressed concerns about the extraordinary capabilities of the Mythos model. However, according to recent developments, the government seems to be encouraging financial institutions to "use poison to fight poison," leveraging the model's exceptional ability to identify vulnerabilities to strengthen banks' defenses.
The Top Banks Have Already Joined In
Although JPMorgan Chase is the only official listed as an initial partner, internal reports indicate that other major Wall Street firms have also joined the testing effort:
Goldman Sachs
Citigroup
Bank of America
Morgan Stanley
Surprisingly, this encouragement comes as Anthropic is currently involved in a legal battle with the Trump administration over being designated as a "supply chain risk" by the U.S. Department of Defense. Tensions had previously escalated due to failed negotiations over restrictions on the government's use of AI models.
Currently, because the Mythos model is too powerful in cybersecurity discovery, Anthropic has restricted access to prevent malicious use. At the same time, UK financial regulators have also begun discussing the potential risks posed by the model. This dynamic marks a shift in AI security regulation from merely "setting barriers" to more complex "collaborative testing."
