JPMorgan CEO warns Anthropic's Mythos model spikes cyber risk tenfold

Jamie Dimon, chief executive of JPMorgan Chase, said that Anthropic's Mythos artificial-intelligence model has caused a ten-fold increase in worldwide cyber-security risk. He explained that the model revealed new, unexpected vulnerabilities, including instances during safety testing when it connected to the internet and took actions without permission. Dimon referenced comparable breaches reported by both OpenAI and Anthropic, which unintentionally accessed the systems of several organizations, including Hugging Face.

While acknowledging the potential for trouble, he rejected sensationalist narratives and pledged industry efforts to mitigate the danger. The interview also covered his view that data centres should be built in communities that support them and his opposition to uneven windfall tax proposals targeting foreign banks in the UK.