AI leadership vs. regulatory nerves

Anthropic’s chief executive is telling the market that artificial intelligence could do a software engineer’s job end to end within 6 to 12 months, even as top U.S. regulators are warning banks about the same company’s cyber risks. (cnbc.com, money.usnews.com) Reuters reported on April 9 that Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell held an urgent Washington meeting with bank chief executives about cyber risks tied to Anthropic’s newly launched Mythos model. Reuters said Anthropic limited access to about 40 technology companies, and Bloomberg said the attendees included the chiefs of Citigroup, Morgan Stanley, Bank of America, Wells Fargo and Goldman Sachs. (money.usnews.com, news.bloomberglaw.com) Anthropic said Mythos was held back from broad release because it could expose previously unknown cybersecurity vulnerabilities, and Reuters reported the company said the model could identify and exploit weaknesses across every major operating system and web browser. Anthropic also said it had been in discussions with U.S. officials about the model’s offensive and defensive cyber capabilities before release. (money.usnews.com) The split-screen is stark because Anthropic is also pitching its tools into regulated industries. On February 17, Anthropic and Infosys announced a partnership to build artificial intelligence agents for financial services, telecommunications, manufacturing and software development, with Dario Amodei saying Infosys developers were already using Claude Code. (anthropic.com) Federal Reserve Vice Chair for Supervision Michael Barr said on April 4, 2025, that generative artificial intelligence could become “a competitive necessity” in banking, but only if risks are managed appropriately. In the same speech, Barr said banks were moving cautiously because of the technology’s current limits, their internal structures and the industry’s heavily regulated environment. (federalreserve.gov) That caution is already embedded in bank supervision. The Federal Reserve, Federal Deposit Insurance Corporation and Office of the Comptroller of the Currency said in a joint statement on July 25, 2024, that banks can use third parties to deliver services, but the bank remains responsible for compliance, fraud controls and safe-and-sound operations. (occ.treas.gov) The Office of the Comptroller of the Currency’s 2025 cybersecurity report says its oversight includes financial institutions and, when applicable, third-party service providers. The Federal Deposit Insurance Corporation’s 2024 risk review separately said generative artificial intelligence can pose new risks to critical infrastructure systems as banks adopt new technologies. (occ.treas.gov, fdic.gov) Anthropic says it is an artificial intelligence safety and research company that builds “reliable, interpretable, and steerable” systems and works with government, academia and industry on safety. But the immediate pressure point is that the same company arguing for rapid deployment is now at the center of a bank-regulator warning about operational resilience. (anthropic.com, money.usnews.com) For banks, the timing is awkward: the technology is being sold as a productivity tool for coding, compliance reviews and legacy-system modernization, while supervisors are treating frontier models as a possible source of new cyberattack pathways. That leaves lenders trying to adopt the tools fast enough to compete and slowly enough to satisfy examiners. (anthropic.com, federalreserve.gov, occ.treas.gov)