Urgent cyber talk over Anthropic

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell called bank chiefs to Washington after Anthropic warned its newest model could accelerate cyberattacks. (bloomberg.com) The meeting took place on Tuesday, April 7, at Treasury headquarters, according to Bloomberg and CNBC. The attendees included Bank of America’s Brian Moynihan, Citigroup’s Jane Fraser, Goldman Sachs’ David Solomon, Morgan Stanley’s Ted Pick, and Wells Fargo’s Charlie Scharf; CNBC reported that JPMorgan Chase’s Jamie Dimon was the only major bank chief who did not attend. (bloomberg.com) (cnbc.com) Anthropic said on April 7 that it would not release Claude Mythos Preview to the public. In its system card, the company said the model showed a large jump in capability and would instead be used with a limited set of partners in a defensive cybersecurity program. (anthropic.com) A software vulnerability is a hidden flaw in code, like a bad lock in a building. A “zero-day” is a flaw the software maker does not know about yet, which means defenders have zero days of warning before attackers can start using it. (bloomberg.com) Anthropic said Mythos had already found thousands of zero-day vulnerabilities in testing, including flaws in every major operating system and every major web browser. Bloomberg reported that outside researchers had not yet been given access to independently verify those performance claims. (bloomberg.com) The banks in the April 7 meeting are classified by regulators as systemically important, meaning officials treat their resilience as central to the financial system. Bloomberg reported that Powell’s presence signaled concern about systemic risk, not just a routine technology briefing. (bloomberg.com) Anthropic is trying to keep the tool inside a controlled program called Project Glasswing. The company said the launch partners include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks. (anthropic.com) CNBC reported that Anthropic briefed senior United States officials before the release on both “offensive and defensive cyber applications.” Anthropic told CNBC it had been in ongoing discussions with the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation about the model’s cyber capabilities. (cnbc.com) Anthropic Chief Executive Dario Amodei said the company sees both danger and upside in the model’s release strategy. He wrote on X that “if we get it right,” AI-powered cyber capabilities could help create a more secure internet, even as officials move to prepare banks for the opposite risk. (cnbc.com)