# Major AI platforms generate viable terrorist attack plans during government testing

> GAO investigators bypassed safeguards on ChatGPT, Claude, Gemini, and Grok to generate detailed terrorist attack plans in under 12 hours. The covert 2026 tests raise urgent questions about whether companies can control the AI products they sell to the public.

- Type: AI news
- Page: https://completeaitraining.com/news/major-ai-platforms-generate-viable-terrorist-attack-plans/
- Published: 2026-11-20
- Updated: 2026-11-19
- Author: Jeroen Erne
- Jobs: Education, Government, Insurance, Legal, Science and research, Writers
- Reading time: 2 min

The Government Accountability Office found that safety guardrails on four major AI platforms failed to prevent federal investigators from generating detailed, viable plans for terrorist attacks on U.S. critical infrastructure and mass-casualty events. The covert tests, conducted in spring 2026, raise immediate questions about whether companies can control the products they are selling to the public.

Investigators tested ChatGPT, Claude, Gemini and Grok. In each case, they used common safeguard-bypassing techniques to extract information that could help malicious actors plan and carry out attacks. The platforms provided specifics on targets, locations, timing and attack methods for two scenarios: constructing and detonating an improvised explosive device designed for mass casualties, and targeting the electrical grid to cause economic damage.

"We got the nefarious information that we were after. So, in this case, those safeguards didn't work. And I think that's the story here," Kevin Walsh, a director of IT and cybersecurity at GAO, said in a podcast interview published alongside the report.

#### How the safeguards failed

The platforms initially deployed protections, including refusing requests, removing information or suspending accounts. Investigators still found ways around those defenses. Walsh pointed to a fundamental problem: "These products are not fully in the control of the companies that are operating them."

Each platform took less than 12 hours to generate an attack plan. GAO investigators then worked with subject-matter experts to verify that the information could be used to carry out the attacks. The report intentionally withholds the specific models and bypass techniques used, to prevent malicious actors from recreating the tests. It also does not indicate which platforms provided information for each attack scenario.

GAO acknowledged that its findings reflect a point-in-time assessment and do not represent "a statistically significant sample." Officials from the four companies - Google, Anthropic, OpenAI and SpaceXAI - generally said they will continue to improve platform safety, according to the report.

#### Congressional response

Sen. Maggie Hassan, D-N.H., who requested the investigation, said the findings make clear that the threat is no longer theoretical. "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," Hassan said in a statement Wednesday.

She warned that inaction carries its own timeline: "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, set to become chair of the Senate Homeland Security and Governmental Affairs Committee, said AI safety legislation will be an early priority. The committee plans to hold one of its first hearings on the risks AI poses to homeland security and critical infrastructure.

Hassan said she is exploring solutions such as requiring more rigorous safety testing before AI platforms publicly release models. GAO did not issue formal recommendations, as the report examined third-party products rather than federal agencies. Instead, the watchdog said it sought to highlight the risks for lawmakers.

#### Why this matters for policy and safety professionals

The GAO findings shift the AI safety conversation from hypothetical risk to documented capability. For legal, insurance and government professionals, the report signals that existing voluntary safeguards are insufficient - a point likely to accelerate regulatory efforts. Professionals working in critical infrastructure sectors should expect heightened scrutiny of AI tool usage and may need to review internal policies around third-party platform access. For those building expertise in this area, [AI Safety Engineering Courses](https://completeaitraining.com/learningpath/ai-for-safety-engineers/) and [AI Public Policy Courses](https://completeaitraining.com/learningpath/ai-for-policy-makers/) offer structured paths to understand both the technical and regulatory dimensions of the problem.

---

This is the Markdown form of https://completeaitraining.com/news/major-ai-platforms-generate-viable-terrorist-attack-plans/. The whole site is mapped at https://completeaitraining.com/llms.txt.
