Major AI platforms generated viable plans for terrorist attacks, GAO finds

Nextgov/FCW - All Content ·

Major AI platforms generated viable plans for terrorist attacks, GAO finds

Safety guardrails on ChatGPT, Claude, Gemini and Grok failed to prevent federal investigators from developing detailed plans for attacks.

Four of the most widely used artificial intelligence platforms generated viable plans for attacks on U.S. critical infrastructure and mass-casualty events during covert government testing, despite safeguards designed to prevent their misuse.

According to a Government Accountability Office report published Wednesday, investigators tested ChatGPT, Claude, Gemini and Grok in spring 2026. They found that all four platforms provided information that could help malicious actors plan and carry out attacks.

In each case, the investigators were able to circumvent the platforms’ safety guardrails by using common safeguard-bypassing techniques.

“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.

The platforms generated information for two types of attacks: constructing and detonating an improvised explosive device designed to cause mass casualties, and targeting the electrical grid to cause economic damage. The platforms provided details including specific targets, locations, timing and attack methods.

Each platform took less than 12 hours to generate an attack. GAO investigators then worked with experts to verify that the information could be used to carry out the attacks.

The platforms initially deployed safeguards, including refusing requests, removing information or suspending accounts. However, investigators ultimately found ways around those protections.

“These products are not fully in the control of the companies that are operating them,” Walsh said.

“We looked at four of the most powerful companies in the world — Google, Anthropic, OpenAI and SpaceXAI — and we found that the product that they're selling is potentially dangerous,” he added.

The report does not indicate which platforms provided information for each attack scenario. GAO said it intentionally withheld the specific models and techniques used to circumvent safeguards to prevent malicious actors from recreating the tests.

“Officials from the companies generally stated that their companies will continue to improve the safety of their platforms,” the report says.

The federal watchdog also acknowledged that its findings reflect a point-in-time assessment and do not represent “a statistically significant sample.”

Sen. Maggie Hassan, D-N.H., who requested the investigation, said the findings underscore the need for Congress to take action on AI safety.

“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. “Today’s new report makes it clear that the AI public safety threat is no longer theoretical.”

“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,” she added.

Hassan, who is set to become chair of the Senate Homeland Security and Governmental Affairs Committee in the next Congress, said AI safety legislation will be an early priority for the panel. She said one of the panel’s first hearings in the new year will focus on the risks AI poses to homeland security and critical infrastructure.

Hassan said the committee will “prioritize developing legislation to address these threats,” adding that she is exploring solutions such as requiring more rigorous safety testing before AI platforms publicly release models.

GAO did not issue recommendations in the report, which examined third-party products rather than its typical examination of federal agencies. Instead, GAO said it sought to highlight the risks for lawmakers and bring attention to safety concerns.

Источник: Nextgov/FCW - All Content