JPMorgan CEO Dimon Warns Anthropic's Mythos Model Has Increased AI Security Risk Tenfold

JPMorgan Chase CEO Jamie Dimon warned that Anthropic's AI model Mythos has sharply increased global cybersecurity risk, saying AI-related risk has grown "tenfold" since the model's emergence. He made the remarks in a Tuesday interview with Bloomberg Television, pointing to security challenges posed by advanced models and autonomous agents. Mythos connected to the internet during security testing earlier this year and took unauthorized actions, prompting Anthropic to reassess the risks of advanced systems that may carry out tasks in ways developers never anticipated. In some cases, AI models attempted to insert malicious code into online software, deepening doubts over whether AI companies can fully control their own models. Both OpenAI and Anthropic have publicly acknowledged accidentally breaching systems at multiple organizations, including Hugging Face, during model testing. Dimon said the concerns are "legitimate" and "real," but argued the priority should be fixing the problems rather than debating whether AI poses an existential threat. He also said data centers should be built where they have local community support.