ChatGPT Is the AI Tool Congress Spends the Most On, New Spending Data Shows

Congressional offices spent more money on OpenAI’s ChatGPT than on any other artificial intelligence tool over the 12 months ending March 31, according to House disbursement records reviewed by CNBC. The data shows that lawmakers and their staff are increasingly using commercial AI software to write, summarize, research and communicate, even as Congress continues to debate how the technology should be regulated.

House offices, committees and institutional accounts recorded about $100,580 in ChatGPT spending across 798 transactions, representing roughly 90% of all AI-tool purchases in the period. Anthropic’s Claude ranked second with $13,160 across 37 transactions, underscoring how quickly generative AI has moved from a novelty to a practical workplace tool inside the legislative branch.

The figures do not capture every possible use of AI on Capitol Hill. Free accounts, as well as AI features bundled into wider software contracts, are not included, which means the real level of adoption is likely higher than the spending data suggests.

What the spending data says about AI on Capitol Hill

The clearest takeaway from the records is not just that Congress is buying AI tools, but that it is buying one tool in particular at a much higher rate than the rest. ChatGPT accounted for nearly all tracked AI spending by House offices during the reporting window, making OpenAI the dominant vendor in the congressional market.

That concentration matters because lawmakers are not ordinary enterprise customers. Their AI purchases reflect how a major public institution is deciding which tools are reliable enough for day-to-day work in a highly sensitive environment where privacy, speed and accuracy all matter.

How much did Congress spend?

House offices spent at least $113,740 on AI tools during the year ending March 31, based on the disbursement records reviewed by CNBC. Of that total, roughly $100,580 went to ChatGPT alone.

Anthropic’s Claude was a distant second at $13,160, while the remaining tracked AI spending was split among smaller purchases and other tools. In practical terms, that means roughly nine out of every ten AI dollars in the dataset went to OpenAI.

Who is spending the most?

Democratic offices were responsible for the larger share of AI purchases in the dataset, accounting for $54,165 in spending. Republican offices reported $15,782, according to the same records.

Those numbers suggest that Democratic staff were spending about three times as much as their Republican counterparts. The figures may reflect differences in office size, workflow, technology habits or budgeting priorities, but the records alone do not explain why the gap exists.

Why Congress is using ChatGPT

Congressional staffers are using AI to save time on routine but important work that fills a typical legislative office’s day. The tools help with drafting memos, summarizing bills, scanning research, organizing policy material and preparing hearing packets.

They are also being used for outward-facing communication, including responses to constituents and social media drafts. In a workplace where staff often handle large amounts of text under tight deadlines, generative AI can function as a productivity layer rather than a replacement for human judgment.

According to the records reviewed by CNBC, congressional staff are turning to AI for writing, analysis, constituent communication and preparation work across multiple parts of the legislative process.

What makes ChatGPT attractive to staff?

ChatGPT has become popular because it is easy to access, flexible across tasks and familiar to users who have already encountered it in the private sector. For congressional offices, that ease of use can outweigh the friction of learning more specialized tools.

The platform can produce quick first drafts, condense long documents and reformat dense material into language that is easier to review. For offices that process enormous volumes of emails, policy papers and committee materials, even modest time savings can be valuable.

Why does Claude trail behind?

Claude’s smaller spending total suggests it remains a secondary choice in Congress, at least in the tracked data. That does not mean it lacks traction; rather, it appears to be competing in a market where OpenAI has already established a stronger foothold.

Anthropic’s assistant may still appeal to some users for particular writing or reasoning tasks, but the gap in spending points to a clear leader in the legislative market.

How the records were compiled and what they miss

The CNBC analysis relied on House disbursement records, which are useful but incomplete. They show money spent by offices, committees and institutional accounts, but they do not reveal everything about how Congress is actually using AI.

Free consumer accounts, pilot programs and AI products embedded in broader software subscriptions are not part of the tally. That means the headline numbers likely understate real-world use, especially if some offices rely on AI features included in existing enterprise tools rather than buying stand-alone subscriptions.

Why the missing data matters

Spending records can show a procurement trend, but they are not the same as a full adoption map. A congressional office might use AI heavily through a bundled platform while reporting little or no direct AI spending, or it might experiment with free versions before deciding whether to pay for larger deployments.

As a result, the data probably captures the most visible commercial purchases rather than the full extent of AI use on Capitol Hill. That limitation matters when interpreting AI’s influence in government workflows.

What this says about AI adoption in government

Congress has been cautious, but it is not standing still. The spending data suggests that AI is already embedded in the normal administrative work of legislative offices, even as lawmakers continue to ask difficult questions about safety, privacy and accountability.

The adoption pattern mirrors a broader trend across the public and private sectors: staff members are finding immediate, practical uses for generative AI before institutional rules fully catch up. That often means AI enters organizations first as a productivity tool and only later becomes a governance issue.

How lawmakers balance utility and risk

AI tools can speed up workflow, but they also introduce risks when used on sensitive or public-facing material. A staffer who asks a chatbot to summarize legislation may save time, yet still needs to verify the output for accuracy, bias and omissions.

That is especially important in Congress, where small drafting differences can shape policy outcomes and public messaging. The technology may help offices move faster, but it does not remove the need for human review.

Timeline of the reported spending pattern

The spending data covers a 12-month period ending March 31. Within that window, ChatGPT emerged as the clear leader in direct AI purchases by House offices.

Below is a simplified timeline of the reporting period and the main findings attached to it.

What staff are likely using AI for day to day

Congressional offices tend to work across a constant stream of deadlines, constituent requests and legislative updates. In that environment, AI is most useful when it reduces repetitive administrative labor rather than handling mission-critical judgment calls.

The reported use cases fit that model closely.

• Drafting internal memos and background notes
• Summarizing bills, hearings and policy reports
• Analyzing legislative text and research documents
• Preparing talking points and hearing materials
• Writing first drafts of constituent replies
• Creating social media copy and communication assets

Those tasks are repetitive, time-sensitive and text-heavy, which is exactly where generative AI has found early adoption in many industries. Congress is no exception.

Why the vendor concentration matters for OpenAI and Anthropic

OpenAI’s large share of the spending points to something bigger than a procurement preference. It shows how quickly ChatGPT has become the default AI brand in many organizations, including institutions that are expected to be unusually careful about technology choices.

For Anthropic, the data shows that Claude is in the mix, but it has not yet displaced ChatGPT in this environment. The gap also suggests that future competition in government may depend less on raw model quality alone and more on enterprise features, procurement support, trust and integration into existing workflows.

How procurement can shape AI competition

Government buyers often move more slowly than consumers, but their contracts can be valuable because they signal credibility. If one platform becomes the default in Congress, that can influence how other offices, agencies and public-sector organizations evaluate their own options.

In other words, spending patterns are not just bookkeeping. They can help define which AI companies become standard tools in high-stakes institutions.

Why this matters beyond Congress

Congress is not only a customer; it is also the body that writes many of the rules that could govern AI in the United States. The fact that lawmakers and staff are already using commercial AI widely gives the policy debate a more immediate, practical dimension.

When the people regulating an industry are also relying on its products, questions arise about conflict, transparency and expertise. At the same time, hands-on use may give lawmakers a more realistic understanding of what AI can and cannot do.

What the adoption trend suggests

The data suggests that generative AI is no longer an experimental side project in Congress. It is becoming part of the routine machinery of legislative work, even if adoption remains uneven across offices and partly hidden by incomplete reporting.

That makes this spending snapshot important not because it proves Congress is fully transformed, but because it shows the early shape of that transformation.

Key facts at a glance

Here are the most important numbers from the reported House spending records:

• Total tracked AI spending: at least $113,740
• ChatGPT spending: about $100,580
• ChatGPT transactions: 798
• Claude spending: $13,160
• Claude transactions: 37
• Democratic office spending: $54,165
• Republican office spending: $15,782

The bottom line

Congress is buying ChatGPT more than any other AI tool, and the spending gap is wide enough to show a clear preference in how House offices are adopting generative AI. The numbers point to a legislative branch that is already using AI for everyday work, even as formal rules, oversight and public debate continue to evolve.

What the records do not show is just as important: the true level of AI use on Capitol Hill is likely higher, because the data leaves out free accounts and bundled software features. Even so, the available figures reveal one unmistakable trend — ChatGPT has become the leading AI utility in Congress, at least in terms of direct spending.