Jamie Dimon, chief executive of JPMorgan Chase (JPMC), has said that the risks associated with advanced AI models such as Anthropic’s Claude Mythos are “a real issue”.

Speaking at the Pennsylvania Defense ​and Innovation Summit, hosted by senator Dave McCormick, Dimon said that actions by the US government to limit the release of Mythos are not to stop small banks from gaining access but to limit the potential damage a widespread release could cause.

“It’s really just to make sure when we roll out this stuff that we know it’s being controlled, because you're giving ballistic missiles to individuals with Mythos basically,” Dimon said.

The longtime JPMC chief, who is reportedly planning his successor after 20 years in the role, was speaking on an “Investing in America” panel that also included US president Trump, defence secretary Pete Hegseth, and Mike Gallagher, head of defence at Palantir.

Dimon’s comments echo growing concerns over the cybersecurity risks advanced AI models pose. On 7 July, the European Central Bank ordered eurozone banks to prepare AI cyber defence plans by 31 October, and agencies in the Five Eyes intelligence alliance have warned that advanced AI attacks are months, not years away.

On 12 June, Anthropic pulled access to Claude Mythos and Claude Fable 5, its most powerful publicly available model, after the US government ordered it to suspend access for foreign nationals over security concerns.

It eventually restored access on 1 July, stating that it had discussed potential risks with the government and introduced new safety measures to prevent the models from being used for malicious purposes.

When first released, Anthropic only made Mythos available to approved organisations within its cybersecurity initiative Project Glasswing. JPMC was a founding partner, alongside technology giants including Amazon Web Services, Apple, Google, and Microsoft as well as cybersecurity firms such as CrowdStrike and Palo Alto Networks.

In the months since its initial release, the firm has provided Japanese megabanks with access to the model and expanded Glasswing to include 150 new organisations across 15 countries.
Many European and UK banks still lack access. Earlier this week Harriet Rees, chief information officer at Starling Bank, told Reuters that the main UK banks with access to the model are regional operations of US lenders and called on Britain to build out its AI capabilities to “protect our leading position moving forward”.

---