A pair of congressional leaders is asking for insight into the use of AI chatbots by President Donald Trump, per a letter sent to the commander-in-chief Thursday.
Reps. Jamie Raskin, D-Md., and Robert Garcia, D-Calif., requested a closed-door briefing with a White House representative to discuss document retention policies governing AI inputs and outputs, vendor contracts supporting use, systems deployed for non-compliance detection, cybersecurity safeguards and data access controls.
The duo, ranking members of the House Committee on the Judiciary and House Committee on Oversight and Government Reform, respectively, also asked for information on any past incidents involving AI chatbots that raised concerns.
“We pose the question because, according to public reports, in December 2025, during an Oval Office meeting with your former advisor, Elon Musk, you ‘spent hours’ talking to Mr. Musk’s AI assistant, Grok,” the congressional leaders said in the letter. “You reportedly sought the chatbot’s advice on carrying out some of your presidential duties, including asking Grok to opine on how Venezuelans would respond to the capture of President Nicolás Maduro.”
Raskin and Garcia cited the U.S. military’s strike on an Iranian elementary school and a near-miss in which the military prepared to intercept a misidentified ship as other examples of AI misuse. The former led to the deaths of 123 children, and the latter, they said, “could have easily precipitated a war with China.”
The congressional leaders requested a briefing by Nov. 11 to give their respective committees time to take appropriate legislative and remedial actions.
As AI creeps into critical operations, its limitations have become a focus of concern. Security flaws, inaccuracies and potential for outputs based on the political or personal agendas of chatbot-makers are top of mind.
Grok, in particular, has faced criticism for its tendency to produce biased and concerning responses.
“Grok has demonstrated persistent and escalating failures related to accuracy, neutrality, and safety — including the generation of racist, antisemitic, conspiratorial, and false content, as well as sexually exploitative outputs,” a coalition of more than 30 civic and public interest groups said in February as part of its third letter to the Office of Management and Budget. “This level of corporate conduct and system failures are patently incompatible with the Administration’s own requirements for federally procured AI systems.”
Despite its reputation, the chatbot continues to gain steam in the federal government. The U.S. Department of Agriculture is sponsoring xAI’s pursuit of a FedRAMP High Authorization. The Department of Health and Human Services, the Treasury Department and the Office of Personnel Management are also tapping the tool to varying degrees.
Elsewhere, the tool has had a harder time building a foothold among workers.
“If the American President is relying on the policy advice of a Musk-designed, Hitler-friendly chatbot with a predilection for child pornography in order to carry out his official duties, Americans have a right to know,” Raskin and Garcia said. “If the United States military killed dozens of innocent schoolchildren based on the faulty advice of an AI chatbot, we have a right to know that too.”
The lawmakers argued that inputs by Trump and his staff, as well as the outputs they received, fall under the Presidential Records Act of 1978. The rule changes the legal ownership of official records of the president from private to public and sets up a structure for document release and management following the end of a presidential term.
Agencies have had to contend with evolving record retention expectations, too.
The National Archives and Records Administration said agencies should retain AI usage and results as a record if they rely on it in decision-making, in conducting official business or when incorporating it into an agency system. But it’s up to agencies to decide when the line is crossed from personal convenience to worthy of a federal record, leaving a gray area ripe for myriad interpretations.
