CRS director: AI can speed tasks like summaries and coding, but research still needs human verification

At a congressional question-and-answer session (date not specified in the transcript), Director Donfried of the Congressional Research Service told a lawmaker that CRS is exploring artificial intelligence but remains committed to producing nonpartisan, authoritative research that members can trust.

"These values are really our promise to you," Donfried said, listing CRS’s five core values—nonpartisan, objective, authoritative, confidential and timely—and stressing the agency’s responsibility to meet members’ deadlines while preserving accuracy and confidentiality. He said the office will prioritize the most urgent information when full, fully sourced memos cannot be produced within a short turnaround.

Asked whether AI could speed CRS’s work, Donfried cautioned that many chatbots pull from the broad internet and do not reliably provide citations. "We're having to verify any analysis that AI is producing," he said, adding that congressional staff often bring chatbot outputs to CRS seeking confirmation: "ChatGPT gave me this answer. Is it true?" Donfried said that, in many cases, the chatbot answers are not accurate and currently do not shorten the time required for authoritative research.

Still, Donfried said CRS is finding practical uses for AI that increase productivity: "It's very good at summarizing meetings and giving you meeting notes. It's very good at drafting emails. It's very good at coding. So we can use those tools to provide code to put graphics into our reports." He emphasized that those efficiencies do not replace the need for human oversight and verification for research conclusions.

The exchange highlighted a trade-off CRS faces when members request rapid guidance: staff may receive targeted facts or a concise answer in a compressed time frame rather than a fully sighted, deeply sourced memo, and CRS will work with requesting offices to determine the most urgent needs.

The session did not record a change in CRS policy but underscored the agency’s cautious approach to integrating AI into research workflows while preserving its statutory independence and standards.