AI executives prepare for a possible catastrophe and public backlash, Axios says
The scenarios include attacks on essential infrastructure and a government response. OpenAI says it does not treat the scenarios in its exercises as inevitable.
Listen to this story
The audio brief
Story brief
3 key pointsThe reported planning goes beyond technical incident response: executives are also weighing how to handle public anger and possible sweeping regulation after a hypothetical AI-enabled attack. OpenAI says it runs preparedness exercises across a range of scenarios and does not regard them as inevitable; Anthropic declined comment. The account offers no evidence of an imminent threat, and the cited six-to-12-month window is an insider forecast, not a company-backed timeline.
- 01
Hypothetical scenarios include attacks on the power grid, water supply, or banking system; no such attacks were reported as having occurred.
- 02
OpenAI said AI is changing the cyber threat landscape and that it is focused on getting capable tools into defenders’ hands.
- 03
The preparations may resemble tabletop exercises long used by institutions such as banks and the Pentagon, rather than a new form of crisis planning.
Executives at OpenAI, Anthropic and other major AI companies are preparing contingency plans for a possible AI catastrophe and the public backlash that could follow, Axios reported October 9. The planning concerns a model causing major public harm. OpenAI confirmed preparedness exercises, while stressing that their scenarios are not treated as inevitable.
The scenarios include an AI model causing major public harm through a hack targeting the power grid, water supply or banking system. These are hypothetical attacks being considered in the planning, not newly disclosed attacks on those services, according to the Axios account cited by the New York Post.
Executives fear such an incident could trigger sweeping government regulations, alongside public anger at AI companies and their leaders. Their preparations include considering potential responses to that political reaction, not just the harm caused by an attack.
Several industry insiders believe a major incident could occur within six to 12 months, according to the Times of India’s account of Axios’s reporting. That is a forecast about a possible crisis, not an established deadline for one or a prediction publicly endorsed by both companies.
OpenAI said its teams discuss and work through a range of potential scenarios in preparedness exercises. A spokesperson described the purpose of those rehearsals in a statement to Axios:
These scenarios are not treated as inevitable, but are meant to help us prepare for a variety of circumstances.
OpenAI spokesperson, speaking to Axios
The company also said AI is changing the cyber threat landscape and that it is focused on getting capable tools into defenders’ hands. Anthropic declined to comment to Axios on the reported preparations.
How unusual these preparations are remains unclear. The New York Post noted that they may resemble tabletop exercises—the scenario-based risk rehearsals long used by institutions including banks and the Pentagon—rather than a fundamentally different kind of crisis planning.
Sources
- timesofindia.indiatimes.com'Inevitable': AI companies preparing for catastrophic event in 6-12 months, claims report
- nypost.comAI companies plot how to respond if catastrophic hacking incident causes ‘revolt’: report
Reader comments
Newest comments first. Replies stay oldest first.