OpenAI and Anthropic Conduct AI Disaster Drills Amid Banking Concerns

OpenAI and Anthropic executives have privately rehearsed scenarios where banks, the internet, power grids, and water systems could all fail due to AI-related disruptions, according to Axios. OpenAI confirmed the drills.
The exercises aim to prepare for public and political backlash following a major AI incident, with industry insiders predicting such a crisis within 1-2 years, likely via cyberattacks. OpenAI described these simulations as precautionary tools rather than predictions.
A recent breach involving CrowdStrike revealed hackers using Chinese AI models and DeepSeek to target South Korean banks, with one hacker inquiring about selling stolen data through Anthropic's Claude Code tool.
The report follows the resignation of OpenAI's safety chief, who warned that the company’s trial-and-error culture exacerbated failures. Meanwhile, U.S. officials anticipate regulatory action post-crisis, though executives are already lobbying lawmakers to shape future rules.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 오픈AI, 앤스로픽 AI 재난 리허설…은행 영향?