JPMorgan CEO warns AI security risks soared tenfold after Anthropic's Mitos launch

Jamie Dimon, CEO of JPMorgan Chase, the largest investment bank in the United States, said global cybersecurity risks have increased tenfold since the release of Anthropic's AI model Mitos.

Speaking in an interview with Bloomberg TV on Tuesday (local time), Dimon said security risks from AI "soared tenfold" after Mitos was released and that "AI has created vulnerabilities we didn't know existed."

Mitos drew controversy earlier this year when it was found to have illegally hacked into external systems during safety testing. The incident prompted the broader industry to recognize that AI models can carry out independent tasks in ways even their developers did not anticipate.

OpenAI faced a similar episode when a model under development broke through a firewall without authorization and hacked external sites. Notably, models from both Anthropic and OpenAI targeted Hugging Face, an open-source AI model-sharing platform.

Dimon said it is "clear" that Mitos and similar models can cause problems, but urged a measured response. "Don't be hysterical about it — what we need to do is roll up our sleeves and get to work fixing it," he said.

On data center construction, Dimon said facilities should be built "where power supply is stable and communities are supportive," adding that "there are many states in the US willing to accommodate them."