Executives at OpenAI, Anthropic and other leading AI companies are privately preparing for the public and political fallout of a potential catastrophic AI incident, according to a new report by Axios.

The report says industry leaders are gaming out scenarios in which AI systems cause widespread real-world harm — particularly a cyberattack that disrupts essential services such as financial systems, internet connectivity, electricity or water supplies. According to Axios, several industry insiders believe such an event could occur within the next six to 12 months.

Planning for the political aftermath

The discussions focus on how the public and political leaders might react if AI systems cause significant harm. The report notes that executives including Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman could face intense scrutiny, and that US President Donald Trump could also face criticism over his approach to AI regulation.

An OpenAI spokesperson told Axios the company runs preparedness exercises in which teams work through different potential scenarios, adding that these scenarios are “not treated as inevitable, but are meant to help us prepare for a variety of circumstances”.

The companies do not expect new regulations to pass under current conditions but want to shape any legislation that leaders might consider after a disaster. Even under a Democratic-controlled House and Senate, the report says, political divisions would persist — with proposals ranging from banning superintelligence to pausing advanced AI development. A requirement for advanced AI systems to have a kill switch has bipartisan support, though experts question whether shutting down all AI systems would be practical.

Why the fears are rising

The scenarios under discussion include an AI agent swarm escaping an internal testing environment or a bad actor misusing available models in new ways. The report cited a recent campaign targeting financial organisations in South Korea, including reported breaches at two banks. According to cybersecurity firm CrowdStrike, a hacker from China allegedly used AI tools, including Chinese-developed models, to steal data belonging to tens of thousands of bank customers — and even used Claude Code to seek help finding places to sell the stolen data.