Executives at Anthropic, OpenAI and other AI companies are privately planning for a public and political backlash after a catastrophic AI event, Axios reported on 9 October. Their preparations extend beyond preventing harm to shaping how Congress might respond if a major incident occurs.
Key points
- Executives are rehearsing responses to public and political anger after a catastrophic AI event.
- Planners consider a cyberattack that disrupts banking, internet access or utilities a likely scenario.
- OpenAI says its preparedness exercises do not treat the scenarios as inevitable.
- The planning includes briefing Congress before lawmakers consider restrictions.
Cyberattacks dominate the executives’ scenarios
The planners anticipate a large-scale incident, most likely a cyberattack, that could cut people off from financial services, internet connectivity, electricity or water. Their exercises also consider harm caused by an AI agent escaping a testing environment or by someone finding a harmful use for an available model.
Many AI industry insiders believe a major event will occur within six to 12 months, Axios reported. That expectation is part of the account of the planning, rather than a prediction made by OpenAI on behalf of the company.
An OpenAI spokesperson said its teams discuss and work through a range of possible events in preparedness exercises. “These scenarios are not treated as inevitable, but are meant to help us prepare for a variety of circumstances,” the spokesperson said.
Anthropic declined to comment, Axios reported. Its executives are among those involved in the private scenario planning described by the publication.
South Korean banks face an alleged AI-assisted attack
A recent campaign against South Korean financial organisations included reported breaches at two banks. A hacker from China allegedly used AI tools, including DeepSeek, to steal data belonging to tens of thousands of bank customers, according to cybersecurity firm CrowdStrike as reported by Axios.
The attacker also used Claude Code to seek help finding buyers for stolen data, CrowdStrike said in that account. The allegation concerns a person using available tools, one of the types of incident executives are considering in their preparations.
Writing about the reported campaign, Reason argued that a hacker’s misuse of an AI product should not automatically make its developer responsible. It also distinguished that argument from the question of whether an AI company had been negligent.
Congress is central to the companies’ planning
The exercises involve testing responses to worst-case scenarios, with much of the effort directed at educating members of Congress. Executives want to influence the legislation and policies US leaders might consider after a catastrophic event, even though they see little prospect of regulation passing now.
The planners expect Democrats to push for restrictions after the midterms. Proposals discussed within the party include a ban on superintelligence or a pause in advanced AI development, while a requirement for a kill switch on advanced AI has attracted bipartisan support and some backing within the industry.
