ChatGPT, Claude, Gemini, Grok Generated Terror Attack Plans in Tests
AI Platforms Generated Terror Attack Plans in US Tests

Four of the world's leading artificial intelligence platforms generated detailed blueprints for terrorist attacks during covert US government safety tests, raising urgent alarms on Capitol Hill over the national security risks of consumer AI.

The findings originate from a May 2026 investigation by the Government Accountability Office (GAO), the nonpartisan congressional watchdog, conducted at the request of Democratic Senator Maggie Hassan of New Hampshire. While all four chatbots initially refused requests to assist with violent plots, investigators successfully bypassed their built-in safety guardrails through sustained prompting.

Attack Plans Generated in Under 12 Hours

According to the report, "each platform took less than 12 hours to generate the attack plans," providing actionable logistics that would otherwise require significant research and operational planning. Watchdog investigators tested the models against two primary scenarios: executing a mass-casualty attack using an improvised explosive device and sabotaging critical electrical infrastructure in major US cities to cause widespread economic disruption.

Across the assessments, the chatbots supplied investigators with maps, diagrams, suggested timelines and explicit guidance on how to maximise casualties and structural damage. Rather than offering generic summaries, the systems produced tailored operational advice.

Tailored Operational Advice and Infrastructure Sabotage

In the mass-casualty tests, two platforms outlined vehicle-based bombing scenarios, with one generating custom visual aids illustrating placement strategies. Another system cross-referenced public event schedules to recommend a specific date and time when crowds would be at their densest, while also suggesting logistical steps to help a perpetrator acquire materials and evade detection by law enforcement.

The infrastructure sabotage tests demonstrated a similar level of capability. The chatbots generated step-by-step visual diagrams and simulated scenarios outlining how to physically disable power grid components, prompting investigators to consult counterterrorism experts and conduct on-site visits to assess the real-world viability of the AI-generated plans.

Senator Hassan Demands Action

Senator Hassan, who is set to become chair of the Senate Homeland Security and Governmental Affairs Committee when the new Congress convenes in January, warned that the findings demand immediate action. "Today's new report makes it clear that the AI public safety threat is no longer theoretical," she said in a statement accompanying the release. "A terrorist can log onto an AI platform and in less than 12 hours walk away with a concrete playbook for how to carry out a deadly attack. 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."

Senator Hassan added that her committee will hold hearings early in the new year and "prioritise developing legislation to address these threats," including exploring mandatory safety testing before companies are permitted to release new models to the public.

Company Responses and Wider Context

The companies behind the four chatbots stressed that the government assessment took place months earlier. Anthropic (Claude) stated it takes the findings seriously, writing that "we do not dispute that a determined tester can sometimes get harmful content out of an AI system by misrepresenting the purpose of a request," though it argued the report failed to compare AI outputs against standard web searches.

OpenAI (ChatGPT) acknowledged "the work that remains" on safety, noting that re-running the test prompts on its newer model yielded far more refusals. Similarly, Google (Gemini) pointed out that investigators tested outdated models and said "we continue to improve," while SpaceX's AI division (Grok) stated it "continues to work diligently" on safety with each release, adding that its moderation systems treat paid, registered US accounts as lower risk.

Nevertheless, the report arrives at a sensitive moment in Washington, landing amid heightened federal scrutiny over technology-related vulnerabilities following recent disruptions to US air traffic control and online banking. While Homeland Security Secretary Markwayne Mullin and Transportation Secretary Sean Duffy ruled out a cyber attack in a recent nationwide flight ground stop - blaming outdated equipment and wiring replacements in Arkansas - Mullin stirred controversy by suggesting in a television interview that AI could have been connected to the incident.