AI companies quietly prepare for major AI disasters

According to Axios, executives from Anthropic, OpenAI and other US AI companies are holding private, closed-door talks about how to handle public and political backlash if a major AI disaster ever happens.

They expect a large-scale incident to happen, most likely a cyberattack, that could knock out financial services, internet communications, or even water and power supplies. If unsafe AI causes real harm for the first time, the public may turn against the technology even more firmly.

Scenario planning is common in business and national security, but many top AI researchers believe a major safety incident is only a matter of time.

An OpenAI spokesperson said the company runs readiness drills to explore possible situations, though it does not see them as certain, and the goal is to help teams stay prepared. Anthropic declined to comment.