US officials urges banks to use Anthropic's Mythos

US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell encouraged bank executives this week to use Anthropic’s Mythos AI model to detect software vulnerabilities, Bloomberg reported.

Institutions including JPMorgan Chase, Goldman Sachs, and Morgan Stanley are among those with access to the system.

The company said access is being limited for now, in part because Mythos is too good at finding security vulnerabilities, even though it was not trained for cybersecurity.

The move comes as governments and regulators explore how advanced AI tools could be applied to financial system security.

The Financial Times reported that UK financial regulators are also discussing the risk posed by Mythos.

🔗 Source: TechCrunch

🧠 Food for thought

Implications, context, and why it matters.

Mythos threatens core software used across the internet

Anthropic’s AI model, Mythos, can find vulnerabilities in nearly every major operating system and web browser with simple prompts 1.
Anthropic says Mythos uncovered some bugs that were up to 27 years old, and their creators or security monitors had missed them 2.
Teams can use the model to help fix holes, yet experts call it “exceptionally dangerous” for attackers and say it could reshape cybersecurity 3.
Canadian bank executives and regulators have held similar “situational awareness” sessions through the Canadian Financial Sector Resiliency Group (CFRG), a public-private body focused on financial-system resilience 3.

US government split over Anthropic puts bank cyber fears up against Pentagon limits

US Treasury secretary Scott Bessent called major US bank chiefs to talk about Mythos-related cyber risk. Federal Reserve chair Jerome Powell also attended, according to reporting 2.
Anthropic is keeping Mythos access limited to a small group of organizations instead of releasing it publicly 2.
In March, Anthropic filed lawsuits over the Department of Defense’s “supply chain risk” label after a dispute about limits on government use of Claude, Anthropic’s main AI assistant, including mass surveillance of Americans and fully autonomous weapons 4.
The fight puts a spotlight on how the US government works with AI companies on ethical red lines and procurement leverage. Researchers from rival AI firms backed Anthropic through an amicus brief (a legal filing from non-parties offering arguments to the court) filed by individuals at OpenAI and Google DeepMind, Google’s AI research division, in their personal capacities 5.