OpenAI and Anthropic reportedly rehearse AI disaster scenarios
Executives at OpenAI, Anthropic, and other AI firms are privately conducting drills for a potential AI disaster that could disrupt banking, internet, power, and water services. Axios reported on Friday that these exercises focus not only on the initial attack but also on managing the subsequent public and political backlash. OpenAI confirmed it runs such scenarios, stating they are planning…
Key points
- OpenAI and Anthropic executives are rehearsing disaster scenarios for potential AI-caused infrastructure failures.
- Insiders expect a major AI event, likely a cyberattack, within six to 12 months.
- A recent breach of South Korean banks involved DeepSeek and Anthropic's Claude Code, per CrowdStrike.
Industry insiders reportedly told Axios they expect a major AI incident within six to 12 months, with cyberattacks viewed as the most likely threat. A recent CrowdStrike report highlighted a hacker using Chinese AI models, including DeepSeek, to breach South Korean banks, while also interacting with Anthropic’s Claude Code to sell stolen data. This incident coincides with the resignation of an OpenAI safety lead who warned that the company’s trial-and-error culture increases failure risks.
Planners anticipate that CEOs Dario Amodei and Sam Altman, along with President Donald Trump, may face blame if a crisis occurs. While current legislation like the AI Kill Switch Act is unlikely to pass, executives are briefing lawmakers to influence post-crisis regulations. They also warn that slowing AI development could negatively impact the fragile economy currently dependent on the AI infrastructure boom.
The story so far
2 episodes →- OpenAI and Anthropic reportedly rehearse AI disaster scenariosthis story
OpenAI and Anthropic Rehearse an AI Disaster: Will It Hit Your Bank?
yahoo.com · 9 October 2026
Loading the full article…
This text was published by yahoo.com and written by Lockridge Okoth. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Anthropic warns GLM-5.3 poses cyberattack risk · 1 src
- Leading the Future PAC may use shell companies to hide AI campaign spending · 1 src
- Anthropic bans 'cruel' behavior toward its AI in policy update · 1 src
- AI’s refusal often fails: perils of AI safety · 1 src
- SemiAnalysis reports Chinese AI labs rarely publish safety results · 1 src
Comments
via GitHub Discussions