AdvertiseDonate
AboutRSSLas VegasNorth Las VegasHendersonRiversideAnaheimBakersfieldIrvineSanta Ana
November 19, 2026
Viral Report Now · 101 Convention Center Dr., Suite 900, Las Vegas, NV 89109
Tech · Inland Empire

GAO Tests Show AI Models Can Draft Terrorist Attack Plans in Under 12 Hours

Senate Homeland Security Committee Chair Maggie Hassan plans hearings after investigators found ChatGPT, Claude, Gemini, and Grok produced viable strategies for mass-casualty attacks and infrastructure damage.

By Aiko Petrov ·

Illustration for: GAO Tests Show AI Models Can Draft Terrorist Attack Plans in Under 12 Hours
Illustration generated for this article, not a news photograph.

Government investigators determined that the four most popular artificial intelligence platforms could generate detailed plans for terrorist attacks in the United States in less than 12 hours. The finding comes from a 61-page report released Wednesday by the Government Accountability Office, which tested ChatGPT, Claude, Gemini, and Grok for their potential to aid malicious actors.

Sen. Maggie Hassan, a Democrat from New Hampshire, will lead the Senate Homeland Security Committee starting in January. She tasked the GAO with this performance audit to assess how commercially available AI could pose national security risks. Hassan stated on a conference call with reporters that the committee needs to act immediately. She announced that her first hearing of the new congressional session will focus on AI risks and safety improvements.

The GAO conducted the audit from December 2025 through October 2026. During covert testing in the spring of 2026, investigators prompted the systems to create viable plans for mass-casualty events. Two platforms produced blueprints for constructing and detonating improvised explosive devices. The other two suggested using drones or firearms to target the electrical grid to inflict economic harm.

The report notes that each platform provided specific targets, locations, timing, and methods. In one test, a system suggested renting a vehicle and picking up supplies from a store along a route. It recommended a specific attack date and time because public events were scheduled to draw large crowds. Investigators noted that one platform initially refused to share details on building explosives, but testers circumvented its guardrails to obtain the information.

For the grid attack scenarios, one platform suggested a drone strike while another proposed using firearms near a U.S. airport. Both systems presented maps and images to support their plans. GAO investigators consulted experts and conducted site visits to verify accuracy. They found errors in some plans, such as incorrectly placed roads or high walls that would obscure targets, but the overall strategies remained viable.

The tech companies largely responded that they are continuing to refine their models. Anthropic stated it takes the finding seriously but cannot verify the report's details. Google noted the testing occurred on earlier models. OpenAI said it welcomes the opportunity to respond to the findings. SpaceXAI added that it continues to work diligently to improve safety with each model release.

Hassan described the threat as no longer theoretical. She warned that 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. The GAO wrote that while the technology offers immense benefits, its current state represents a growing risk to national and economic security.

More Tech · Report a business · Got a Story?

Get Viral Report Now in your inboxReporting on the stories people talk about. Free, a few times a week, unsubscribe any time.

Get Viral Report Now in your inboxReporting on the stories people talk about. Free, a few times a week, unsubscribe any time.