Banks pressed on Anthropic Mythos
U.S. officials have flagged Anthropic’s Mythos model as a potential cyber risk and Treasury Secretary Scott Bessent summoned bank executives for briefings on the issue. (thehill.com). Banks including Goldman Sachs and Citigroup are reportedly testing Mythos to detect vulnerabilities after the warning, and reporting suggests thousands of zero‑day issues have already been uncovered in tests. ( )
Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell called top bank chiefs to Washington this week after Anthropic’s new Mythos model raised cyber alarms. (thehill.com) The meeting took place at the Treasury Department on Tuesday, April 7, while executives were already in town for the Financial Services Forum, according to The Hill and CNBC. Bank of America’s Brian Moynihan, Citigroup’s Jane Fraser, Goldman Sachs’s David Solomon, Morgan Stanley’s Ted Pick and Wells Fargo’s Charlie Scharf attended; JPMorgan Chase’s Jamie Dimon did not. (thehill.com; cnbc.com) Officials did not point to a specific attack on banks, but they told executives to take Mythos seriously and use it to look for weaknesses in their own systems, Bloomberg reported on April 10. Goldman Sachs, Citigroup, Bank of America and Morgan Stanley are among the banks now testing the model internally, with JPMorgan Chase already named as a launch partner. (bloomberg.com; cnbc.com) A zero-day is a software flaw the developer does not know about yet, which means there is no patch when it is first found. Anthropic said Mythos has identified thousands of zero-day vulnerabilities, including flaws in major operating systems, web browsers and other widely used software. (anthropic.com; techcrunch.com) Anthropic released Mythos on April 7 as a preview, not a public product. The company said the model would be limited to a small set of partners for defensive cybersecurity work because its capabilities were too risky for broad release. (anthropic.com; thehill.com) The company wrapped that rollout into a program called Project Glasswing. CNBC reported that JPMorgan Chase joined technology companies including Apple, Google, Microsoft and Nvidia in the initiative, while TechCrunch reported Anthropic said 12 partner organizations would use the model for defensive security work. (cnbc.com; techcrunch.com) Anthropic said it briefed United States officials before the release on Mythos’s “offensive and defensive cyber capabilities.” CNBC reported those talks included the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation. (thehill.com; cnbc.com) The backdrop is a wider fight between Anthropic and the Trump administration over national security and AI guardrails. The Hill reported in March that the Pentagon moved to cut off government use of Anthropic and label the company a supply-chain risk after disputes over safeguards on military and surveillance uses. (thehill.com; thehill.com) Anthropic’s position is that restricting Mythos now reduces the chance that criminals or spies get a tool that can find bugs faster than human teams can. The banks’ position, as described by officials and people familiar with the testing, is that the same tool may now be needed to check whether their own defenses can hold. (anthropic.com; bloomberg.com)
