Incoming U.S. Senate Homeland Security Committee chair Maggie Hassan says she will hold hearings on AI after government watchdog testing found that widely used AI platforms could generate proposals for terrorist attacks in the United States when prompted by investigators. Hassan, a Democrat, says she has asked the Government Accountability Office to examine how commercially available AI can be misused and what risks it creates for national security.
The GAO report described covert testing in which four major AI platforms—identified as ChatGPT, Claude, Gemini and Grok—were able within about 12 hours to produce “viable” plans. The report says the systems could help outline scenarios involving mass-casualty attacks using explosive devices in U.S. cities and could also be used to target parts of the electrical grid for economic harm. Hassan frames the issue as a near-term public safety and security concern, saying the threat is no longer theoretical.
The reporting also notes the companies’ responses included in the GAO document. Providers including Anthropic, Google, OpenAI and SpaceXAI said they continue to refine safeguards as AI capabilities evolve, while the GAO warns that misuse risk is growing in the current state of the technology.