OpenAI captured roughly 90% of all House spending on AI tools during the year ending March 31, 2026, according to House disbursement records analyzed by CNBC and reported by TechCrunch. Congress spent approximately $100,580 across 798 transactions on ChatGPT, out of at least $113,740 in total AI spending by House offices, committees, and institutional accounts (TechCrunch).

Anthropic's Claude ranked second at $13,160 across 37 transactions. No other AI tool registered meaningful spending in the disclosed records. The gap between first and second place is wide enough that the question is less which vendor Congress prefers and more whether any competitor is in the conversation at all.

Democratic offices accounted for $54,165 in AI purchases, roughly three times the $15,782 spent by Republican offices. The remainder came from committees and institutional accounts rather than individual member offices. The spending figures exclude AI adoption through free accounts or AI functionality bundled into broader software contracts, so the true footprint of AI use on Capitol Hill is larger than these numbers suggest (TechCrunch).

Congressional staffers use ChatGPT and other AI tools to write memos, summarize and analyze legislation, respond to constituents, prepare hearing materials, sort through policy research, and draft social media posts. In other words, the use cases span nearly the full range of routine staff work, from constituent communication to legislative analysis.

The House of Representatives publishes receipts and expenditures for representatives, committees, leadership, and officers, along with disclosures covering income, foreign travel, privately reimbursed travel, and legal expenses (house.gov). The AI spending figures surfaced through this disbursement transparency mechanism, which makes individual transactions searchable but does not aggregate them by vendor or category.

Separately, OpenAI deployed a combined 324 lobbyists during the second quarter of 2026, according to lobbying disclosure analysis by Issue One (Issue One). That figure covers lobbying activity across all levels of government and is not limited to congressional advocacy, but it provides context for the scale of OpenAI's public-sector engagement. The company is simultaneously the dominant vendor selling AI tools to Congress and one of the most aggressive spenders on lobbying to shape the regulatory environment around those tools.

The spending data invites several observations. The concentration of purchases on a single vendor creates a practical dependency: if OpenAI changes its pricing, terms of service, or model behavior, House offices have limited alternatives already in their procurement pipeline. Claude's presence in the records indicates some offices are exploring alternatives, but at less than one-seventh the transaction volume, that exploration is nascent.

The partisan spending gap raises its own questions. Whether the disparity reflects differing attitudes toward AI adoption, differences in office budgets, or simply earlier procurement decisions by Democratic staff cannot be determined from disbursement records alone. The data shows where the money went, not why.

What the figures do establish is that AI tools have moved from novelty to line item in congressional operations within a remarkably short window. Staffers are not experimenting with these tools in isolation; they are spending real, if modest, budget on them across hundreds of transactions, integrated into daily workflows that touch everything from constituent correspondence to legislative analysis.

The total dollar amount, $113,740 across the entire House over twelve months, is modest by enterprise software standards. A single mid-size private-sector firm would likely spend more on AI tooling. But Congress is not a private firm. It is the body that will write the laws governing AI deployment, and its staff are now daily users of the products they may be called upon to regulate, with the dominant vendor among the largest spenders on influence operations in Washington.

That is not inherently a problem. But it is a fact worth holding in view.