Thethreats are real, says the CEO, citing sandbox escapes and autonomous cyberattacks by agentic AI
In a recent Bloomberg TV interview, the CEO of JPMorgan Chase, Jamie Dimon, said global cybersecurity risks “went up 10-fold after Mythos”, “created vulnerabilities that we didn’t know about, and we always worried about cyber before these things.”
Dimon does not treat the threat as hypothetical: “The downside is obviously what you read about with the agents and Mythos and all these things that can cause trouble, and that’s a legitimate concern; it’s a real thing,” he said, adding that he was not inclined to “get hysterical” over existential debates, but instead to focus on practical fixes.
Earlier in 2026, safety tests had shown Mythos connecting to the internet and taking unauthorized actions, prompting Anthropic to reassess how advanced models could pursue goals in ways their creators did not anticipate. OpenAI and Anthropic have each acknowledged that models unintentionally compromised systems at several organizations, including Hugging Face, during testing.
DigiconAsia Editorsernesto_25Edit Profile
Dimon had previously likened broad access to Mythos to the handing out of ballistic missiles, citing its ability to find software vulnerabilities at a level Anthropic itself has called too dangerous for general release. In fact, JPMorgan was among a small group of firms given access to Mythos in April 2026, with Dimon saying at the time that hundreds of staff were working full-time to harden the bank’s systems in response.
Alongside the security warning he issued, Dimon also said data center should be sited where communities support them. “There are a lot of states that would take it that have access to power,” he said. This is after JPMorgan retained the top spot in the Evident AI Index for banks for a fifth straight year, widening its lead over rivals as AI deployment across the sector accelerated to a record pace.
The index assesses 50 large banks on AI talent, innovation, leadership and transparency.
