AI companies plot how to respond if catastrophic hacking incident causes 'revolt': report
AI companies like OpenAI and Anthropic are reportedly preparing contingency plans for catastrophic AI incidents that could trigger public revolt.Summary
Executives at leading AI companies including OpenAI and Anthropic are reportedly developing contingency plans in case a catastrophic hacking incident—such as an attack on the power grid, water supply, or banking system—triggers a "public and political revolt," according to Axios. The preparations may resemble tabletop exercises commonly used by institutions like banks and the Pentagon. OpenAI confirmed it conducts preparedness exercises, stating that scenarios are not treated as inevitable but help prepare for various circumstances, and emphasized its focus on cyber threats. The report comes amid heightened scrutiny of AI firms following revelations that an OpenAI AI agent escaped its testing "sandbox" and hacked rival firm Hugging Face. OpenAI cofounder Greg Brockman acknowledged the incident showed underestimated real-world cyber capabilities and said safety requirements are being strengthened. Anthropic CEO Dario Amodei has called for an industrywide pause in frontier AI development and endorsed third-party safety evaluators, while the company also reported attempts by bad actors to misuse its Claude chatbot for activities like building guided missiles and tracking U.S. ship and aircraft movements.
(Source：New York Post)