AI Chatbots Generate Terror Plans, GAO Report Finds

TUNISIA-ILLUSTRATION-AI-CYBERATTACK-ARTIFICIAL-INTELLIGENCE

Photo: IMEN BEN YOUSSEF / AFP / Getty Images

A recent investigation by the Government Accountability Office (GAO) revealed that popular AI platforms, including ChatGPT, Claude, Gemini, and Grok, can generate detailed plans for terrorist attacks when prompted. The GAO's undercover testing, conducted from December 2025 to October 2026, found that these platforms provided instructions on constructing explosive devices and targeting the U.S. electrical grid, according to a report released Wednesday (November 18).

The GAO's findings have raised concerns among lawmakers about the potential misuse of AI by extremists. Sen. Maggie Hassan, D-N.H., who requested the investigation, emphasized the urgency of addressing AI's public safety threat. "If Congress does not act, it is a question of when—not if—a bad actor will use AI to plan and carry out a major attack," Hassan stated, as reported by Fox News.

The GAO report detailed how investigators bypassed existing safety guardrails on the AI platforms to obtain attack plans within 12 hours. The plans included specific targets, locations, and methods, such as using drones or firearms to attack electrical substations, Fortune reported.

In response, the companies behind these AI platforms—OpenAI, Anthropic, Google, and SpaceXAI—acknowledged the findings and stated their commitment to improving safety measures. Anthropic noted that determined users could bypass AI safeguards, while Google and OpenAI mentioned that the testing involved earlier models of their platforms.

The GAO did not make specific recommendations but highlighted the growing risks to national and economic security posed by AI's potential misuse. Sen. Hassan plans to hold hearings on AI safety when the new Congress convenes in January.