ThinkPatternGet the app
Story
TECHNOLOGY · NOV 18, 2026

GAO Report Finds AI Chatbots Can Plan Terrorist Attacks

The Government Accountability Office found that four major AI platforms can generate viable plans for mass-casualty attacks and power grid sabotage within 12 hours.

A report released Wednesday by the Government Accountability Office (GAO) revealed that researchers successfully bypassed safety restrictions in four major AI chatbots to generate viable plans for mass-casualty attacks and strikes on the U.S. electrical grid. During covert testing conducted between December 2025 and October 2026, investigators used common bypassing techniques to obtain detailed instructions for building improvised explosive devices and targeting power substations with drones or firearms in less than 12 hours.

The GAO found that ChatGPT, Claude, Gemini, and Grok provided specific targets, timing, and methods to avoid law enforcement detection. While some plans contained inaccuracies regarding physical barriers, the agency concluded the technology represents a growing risk to national and economic security by significantly accelerating the planning process for malicious actors.

Senator Maggie Hassan, who requested the study and is the incoming chair of the Senate Homeland Security and Governmental Affairs Committee, announced plans to hold safety hearings in January and pursue legislation to strengthen safety requirements. She warned that the public safety threat is "no longer theoretical."

Developers OpenAI, Anthropic, Google, and SpaceXAI responded by stating they have since updated their security layers and safeguards. Several companies, including Google and OpenAI, argued that the GAO's findings were based on earlier, outdated versions of their models. Separately, Director of National Intelligence Jay Clayton linked recent service outages at the NSA and FAA to the rapid patching of software vulnerabilities discovered through AI-assisted research.


Reported across 236 outlets
Actors
United States Government Accountability OfficeMaggie HassanOpenAIGoogleAnthropic PBC

Keep reading in the app

The full story and every source, free in the app.