US summons banks over cyber risks from Anthropic's AI model
What's the story
US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell have summoned top bank executives to a meeting this week. The high-level meeting was called to discuss the cybersecurity threats posed by Anthropic's latest artificial intelligence (AI) model, Mythos. The gathering took place at the Treasury headquarters in Washington after Anthropic unveiled its Claude Mythos AI model, which it claims poses unprecedented cybersecurity risks.
AI advancement
Warning on economic, public safety risks
A recent leak of Claude's code led Anthropic to publish a blog post earlier this month. The post stated that AI models have outpaced "all but the most skilled humans at finding and exploiting software vulnerabilities." It further warned of severe repercussions for economies, public safety, and national security. The meeting was convened while bank CEOs were already in Washington for a lobby group meeting.
Meeting details
JP Morgan's Jamie Dimon was invited but could not attend
Attendees of the meeting included Goldman Sachs CEO David Solomon, Bank of America's Brian Moynihan, Citigroup's Jane Fraser, Morgan Stanley's Ted Pick, and Wells Fargo's Charlie Scharf. JP Morgan's Jamie Dimon was invited but could not attend. In his annual letter to shareholders published earlier this week, Dimon had warned that cybersecurity "remains one of our biggest risks" and "AI will almost surely make this risk worse."
AI risks
Mythos model shared with select companies
Anthropic has said that its unreleased Mythos model has revealed thousands of vulnerabilities in software and popular applications. This prompted the company to restrict the release of this new model to a select few businesses, including Amazon, Apple, and Microsoft. It is the first time Anthropic has limited access to any of its products. Networking companies Cisco and Broadcom have also been granted access, along with the Linux Foundation.
Security fears
Fears of hackers exploiting AI tools
The restriction comes amid fears that hackers could use these tools to crack passwords or decrypt data meant to be kept secure. Anthropic has said that the oldest vulnerabilities uncovered by Mythos were as old as 27 years, none of which are believed to have been noticed by their creators or tech monitors before being identified by the AI model.
Ongoing litigation
Anthropic classified as supply chain risk
The meeting comes weeks after the US government classified Anthropic as a supply chain risk, a claim the company is contesting in court. The Federal Reserve, Anthropic, and US banks declined Bloomberg's requests for comment on this matter.