Executives at OpenAI and Anthropic are privately planning for a catastrophic AI-enabled cyberattack on banks, power grids or water systems, with many insiders telling Axios they expect such an event within six to twelve months. The planning is focused on Congress, not on households that would lose access to essential services.