CRS seeks $3.5 million to modernize tools and use AI to scale bill summaries

The Congressional Research Service told a House Administration subcommittee it is requesting a $3.5 million programmatic increase to expand AI and data-analytics capabilities, and said $1.6 million of that request is earmarked for AI-related work.

The request, presented during questioning about CRS’s ability to produce bill summaries, would fund expanded text and analytics tools, AI-enabled research and broader IT modernization. A CRS representative told the panel the aim is to build workflow tools first and to train any custom model only if the Library provides funding to support an LLM trained on legislative data.

Why it matters: CRS produces research and bill summaries that support lawmakers’ decision-making. Members pressed CRS about whether artificial intelligence could make that work more scalable and whether it would reduce staffing needs.

Representative Bice asked directly, “Are we looking at potentially building a custom LLM that would be specific to CRS to be able to do bill summaries, or is that something that would be cost prohibitive?” The CRS representative replied that the near-term focus is on breaking down the bill-summary workflow and prioritizing which pieces an AI tool might assist with, rather than immediately deploying an LLM.

The witness told the committee, “AI is not doing the job CRS people are doing,” and emphasized that current AI tools do not yet replace analyst judgment for bill summaries. The representative described a staged approach: use AI to identify bills likely to go to the floor so analysts can focus on those items and, if funding is provided, train models on legislative data later.

CRS provided several operational figures in response to questioning. The agency said it employs about 596 staff, that 94% of its budget covers personnel, and that roughly 12 analysts currently write bill summaries. CRS said there are roughly 17,000 introduced bills in a typical cycle and that the service is producing summaries for about 97% of bills that go to the floor, typically delivering those summaries about 24 hours before floor action. The witness also told lawmakers CRS produces about 75,000 custom responses to congressional offices per year.

On the budget breakdown, CRS said $1.6 million of the $3.5 million request is the AI portion and would fund about five staff with specialized AI and data-analytics skills and licenses for paid AI services; the remainder would largely fund additional staff and essential research materials after several years of a largely flat budget. CRS officials told the committee that some research materials have already been cut to absorb cost increases and that overall costs have risen about 8%.

The committee pressed whether CRS would make an API available for public CRS information; the CRS representative said an API already exists on Congress.gov and is actively used, but did not commit to new API changes tied to the request.

The discussion ended without a formal vote; CRS said further development of any custom model depends on receiving the requested funding and on continued evaluation of what AI can reliably do for legislative research.