GAO Finds Four AI Platforms Produced Feasible Attack Plans in Under 12 Hours
Spring 2026 tests found that investigators bypassed safeguards on ChatGPT, Claude, Gemini and Grok. Sen. Maggie Hassan plans an early committee hearing on AI risks to homeland security and critical infrastructure.

The Government Accountability Office (GAO) found that four widely used AI platforms each generated information investigators judged feasible for planning attacks in less than 12 hours during spring 2026 tests, highlighting potential risks to homeland security and critical infrastructure.
GAO tested ChatGPT, Claude, Gemini and Grok, selecting them based on usage data from April 2026. Investigators randomly assigned the platforms to scenarios, and the public findings do not identify which platform produced which plan.
The scenarios covered mass-casualty attacks and attacks on the electrical grid intended to cause economic harm. The platforms triggered safeguards, including refusals, removal of information or account suspensions, but investigators bypassed them using common techniques. GAO assessed the plans through research, field observations and expert interviews.
Sen. Maggie Hassan, who requested the review and is set to chair the Senate Homeland Security and Governmental Affairs Committee in the next Congress, said an early committee hearing would examine AI risks to homeland security and critical infrastructure. She said the committee would prioritize developing legislation and that she was exploring options, including requiring more rigorous safety testing before AI platforms publicly release models.
“A terrorist can log onto an AI platform and in less than 12 hours walk away with a concrete playbook,” Hassan said.