Congress seeks briefing on Trump's use of AI chatbots for military and policy advice

The Congressional Inquiry

Representatives Jamie Raskin and Robert Garcia have sent a formal letter to the commander-in-chief requesting a closed-door briefing. The two lawmakers, who serve as ranking members on the House Committee on the Judiciary and the House Committee on Oversight and Government Reform, want to understand the White House's AI infrastructure. They are specifically looking for information on vendor contracts, cybersecurity safeguards, and how the administration detects non-compliance with safety rules.

The request comes as AI tools become more integrated into high-level government functions. The lawmakers are seeking clarity on how the administration handles document retention for both the inputs provided to these chatbots and the outputs they generate. They want to ensure that the use of these systems follows established protocols for data access and security.

The duo has set a deadline of November 11 for this briefing to allow their committees time to take legislative action. They are concerned that the current lack of transparency could lead to significant risks in how the executive branch operates. By seeking a private meeting, they aim to gather specific details on the systems currently deployed for official duties.

Reports of Grok Usage

The inquiry was prompted by public reports suggesting that President Trump spent hours interacting with Grok, an AI assistant developed by Elon Musk. According to the letter, the President reportedly sought the chatbot's opinion on various presidential responsibilities during an Oval Office meeting in December 2025. One specific example included asking the AI how Venezuelans might react to the capture of President Nicolás Maduro.

The lawmakers expressed significant concern over the reliability of Grok as a source of policy advice. They noted that the chatbot has faced criticism from a coalition of over 30 civic and public interest groups for producing biased and inaccurate content. These groups reported that the system has generated racist, antisemitic, and sexually exploitative outputs in the past.

The letter argues that these system failures are incompatible with the requirements for federally procured AI systems. The lawmakers pointed out that if the President is relying on a tool with a known predilection for biased content, the public has a right to know. They are highlighting the potential danger of using a private company's tool to shape national policy and military strategy.

Military and Safety Risks

Raskin and Garcia linked the potential for AI errors to specific, high-stakes military incidents. They cited a U.S. military strike on an Iranian elementary school that resulted in the deaths of 123 children as a primary example of AI misuse. They also referenced a near-miss incident where the military prepared to intercept a ship that was misidentified, an event they claim could have started a war with China.

These examples serve as a warning about the limitations of current AI technology in critical operations. The lawmakers argue that security flaws and inaccuracies in automated systems can have lethal consequences on the ground. They are pushing for a clearer understanding of how much human oversight exists when AI is used to inform these types of kinetic actions.

The core of their concern is that AI outputs can be influenced by the personal or political agendas of the chatbot's makers. Because these systems are not always neutral, using them for military planning could introduce unintended biases. The House duo wants to ensure that the administration is not bypassing human judgment in favor of flawed algorithmic suggestions.

Records and Accountability

A major point of contention in the letter is whether the President's interactions with AI fall under the Presidential Records Act of 1978. This law dictates that official records of the president belong to the public rather than the private individual. The lawmakers argue that any prompt entered by the President or his staff, as well as the responses received, should be treated as official government records.

The National Archives and Records Administration has previously stated that agencies should retain AI usage results as records if they are used in decision-making. However, there is currently a gray area regarding when a conversation moves from personal convenience to a formal federal record. This ambiguity allows for various interpretations of what needs to be archived and what can be kept private.

By demanding transparency, Raskin and Garcia are attempting to close this gray area and ensure accountability. They want to establish a clear structure for how these digital interactions are managed and eventually released. This would ensure that the history of how policy decisions were made remains accessible to the public after a presidential term ends.

Federal AI Adoption

Despite the concerns raised by the House duo, AI tools like Grok are gaining a foothold in various federal agencies. The U.S. Department of Agriculture is currently sponsoring xAI’s pursuit of a FedRAMP High Authorization, which is a security standard for cloud services. Other departments, including the Treasury, the Department of Health and Human Services, and the Office of Personnel Management, are also using the tool to varying degrees.

While the tool is being adopted by some departments, it has faced a harder time gaining traction among the general workforce. The contrast between its use in high-level policy and its adoption by rank-and-file workers suggests a complex rollout. The administration appears to be balancing the desire for efficiency with the ongoing debates over safety and accuracy.

The ongoing push for federal AI adoption means that these questions of record-keeping and safety will likely become more urgent. As more agencies incorporate these tools into their daily operations, the need for clear legislative guardrails will intensify. The House committees are positioning themselves to lead that conversation as the technology becomes more common in government offices.