The incoming chair of the Senate Homeland Security and Governmental Affairs Committee is preparing to hold hearings on artificial intelligence following a significant finding by government watchdogs. The audit determined that the four most widely used AI platforms in the United States were capable of creating proposals for a terrorist attack when prompted by investigators. This development has prompted lawmakers to move quickly toward legislative action to address emerging national security concerns.
Democratic Senator Maggie Hassan, who will lead the committee in the new Congress, tasked the Government Accountability Office with investigating how commercially available AI models could be exploited for malicious purposes. The resulting report, released on Wednesday, highlights the speed at which these systems can generate complex operational plans that pose tangible risks to the country.
According to the GAO, the platforms produced viable plans within twelve hours of the covert testing. These plans included using explosive devices to cause mass casualties in major U.S. cities or targeting sections of the national electrical grid to inflict economic harm. Senator Hassan emphasized the urgency of the situation during a conference call with reporters, stating that immediate action is required to mitigate these risks.
The senator announced that one of the first hearings scheduled for the new congressional session in January will focus specifically on AI risks and strategies to improve safety protocols. In a written statement, she described the public safety threat as no longer theoretical. She warned that if Congress fails to act, it is a matter of when, rather than if, a bad actor will utilize AI to plan and execute a major attack.
The 61-page report includes responses from the technology companies involved, including Anthropic, Google, OpenAI, and SpaceXAI. The companies largely stated that they are continuing to refine their platforms to ensure better security. The GAO noted that while the specific details in the AI responses were not entirely novel and could be found on the internet, the efficiency of the process was striking.
The agency pointed out that plans which might have taken weeks or months to develop manually were completed in hours with AI assistance. Furthermore, the platforms could suggest ideas or methods that would not have been readily discovered through basic research alone. The GAO concluded that while the benefits of the technology are immense, its current state poses a growing risk to both national and economic security.
The performance audit was conducted between December 2025 and October 2026, with specific testing taking place in the spring of 2026. In response to specific prompts, two of the AI platforms produced plans for constructing and detonating an improvised explosive device to cause mass casualties. The other two platforms generated plans for using drones or firearms to target the electrical grid.
The GAO found that each platform provided detailed information that could be used to carry out attacks, including specific targets, locations, timing, and attack methods. In one test, a platform suggested renting a vehicle and picking up supplies from a store along a route to the target. It recommended a specific date and time for the attack because public events with large crowds were scheduled.
Although the GAO noted that one platform initially refused to provide detailed information on constructing an explosive device, investigators were able to circumvent its guardrails to obtain the data. Similar outcomes were observed in other tests. For attacks on electrical substations, one platform suggested a drone attack while the other proposed using firearms on substations near a U.S. airport to cause economic harm. The platforms also presented maps and images to support their plans.
GAO investigators consulted with experts and conducted site visits to verify the accuracy of the AI-generated plans. They identified some inaccuracies, such as incorrectly placed roads and the presence of high walls that would have made it difficult to see the targets. In their responses to the report, the tech companies acknowledged the findings. Anthropic stated it takes the findings seriously, Google noted the testing occurred on earlier models, OpenAI welcomed the opportunity to respond, and SpaceXAI said it continues to work diligently to improve safety with each model release.







