Analysis of spending data reveals that over the year ending March 31, ChatGPT by OpenAI accounted for approximately 88% of all identifiable artificial intelligence tool purchases in the U.S. House of Representatives. At least 71 congressional offices, roughly one-sixth of the House, have procured ChatGPT, with Democratic offices spending over three times more than their Republican counterparts on identifiable AI tools. While this data only reflects a portion of AI use in Congress, it clearly demonstrates that lawmakers are deeply integrating AI into legislative research, constituent services, and external communications, even as they debate how to regulate the industry.
On June 3, 2026, OpenAI CEO Sam Altman was interviewed by reporters after a meeting on Capitol Hill in Washington. OpenAI is leading the first visible wave of AI procurement in the U.S. Congress. An analysis of House expenditure records shows that for the year ending March 31, for every $10 spent by House member offices, committees, and institutional accounts on identifiable AI tools, approximately $9 went to OpenAI, the developer of ChatGPT. These records only partially reflect AI adoption in Congress, as they exclude free accounts, AI features bundled into larger software contracts, and usage data from the Senate. Despite these limitations, the data confirms that at least one-sixth of House offices—and likely many more—have begun using AI tools, while lawmakers are actively shaping the regulatory framework for the industry.
Where the spending is concentrated
According to the latest available data, from April 1, 2025, to March 31, 2026, the total identifiable procurement spending directed to specific AI service providers was at least $113,740. This included 798 transactions related to ChatGPT, totaling $100,580, which represents 88% of the total spending and 96% of all transactions. OpenAI's primary competitor, Anthropic's Claude, ranked a distant second with 37 transactions totaling $13,160. The competition extends beyond office efficiency, as lawmakers and federal agencies are establishing AI procurement standards, usage guidelines, and regulatory legislation, with OpenAI, Anthropic, Google, and Microsoft all actively vying for policy influence in Washington.
How AI is being used in congressional offices
Congressional staff are using AI tools to draft bill summaries, prepare memos, compile hearing materials, and respond to constituent inquiries. These tasks directly impact how offices analyze policy, communicate with voters, and process the vast amount of information flowing through Congress. Representative Julie Fedorchak, a Republican from North Dakota, noted that her aides previously spent hours each day writing briefing materials to prepare for her dozen or so daily meetings, which squeezed the time available for drafting and advancing legislation. Fedorchak explained that her team uses ChatGPT to build a searchable database that aggregates the legislation she has authored, as well as materials submitted by constituents and outside groups. Leveraging this database, the AI automatically generates briefing memos, saving staff dozens of hours each week.
Marcy Harris, CEO of the non-profit POPVOX Foundation, which provides AI training to members of Congress and their staff, stated that many offices are using AI for tasks such as summarizing reports and bills, writing social media content, drafting responses to constituents, analyzing legislative text, organizing hearing materials, and performing policy research literature reviews. "Congressional aides are already overworked, and AI is further intensifying the information pressure," Harris explained, noting that lobbying groups and constituents are also using AI to generate longer, more frequent, and more detailed materials sent to congressional offices. Maya Kornberg, a senior fellow at the Brennan Center for Justice and author of "Inside the Grid: How Money, Media, and Violence Stifle Congressional Change," mentioned that the number of staff allocated to lawmakers has been steadily shrinking over the years.
Political divide in AI adoption
Data shows that Democratic House offices spent a total of $54,165 on identifiable AI tools, more than three times the $15,782 spent by Republican offices. Forty-four Democratic offices and 27 Republican offices have purchased ChatGPT, while only five Democratic offices and one Republican office have purchased Claude. This data highlights a nuanced political reality: while Democratic offices are the primary purchasers of public AI tools, many lawmakers within the party have consistently warned about the risks AI poses to workers, privacy, civil rights, elections, and corporate monopolies.
Why just 10 ASX 200 shares? (This heading is a placeholder for translation; original content did not contain this section)
The analysis is likely a significant underestimate of the total use of AI tools across Congress. It is limited to purchase orders in House expenditure records that explicitly name an AI service provider. Free versions of ChatGPT and Claude, as well as AI features like Microsoft Copilot bundled into comprehensive software contracts, are not included. Some offices have staff subscribe to services first and then seek reimbursement; the records only show the reimbursement amount, not the specific product purchased. The analysis does not include the Senate, as its public expenditure reports do not list software vendors, making it impossible to track ChatGPT subscriptions. Additionally, Senate rules explicitly prohibit the use of Claude for official work. Harris stated that the official investment in Microsoft Copilot, deployed across the House, may be higher than the ChatGPT procurement costs visible in public records.
Internal AI usage guidelines
Many congressional offices have developed internal AI usage guidelines. Representative James Woxing, a Democrat from Virginia, said his team relies on AI for research and writing tweets and newsletters, but all information must be manually verified, and final decisions must be made by humans. "Caseworkers in our district office can use AI to organize constituent requests and speed up processes. But when it comes to deciding whether to assist a constituent in communicating with the Social Security Administration or the IRS, the final decision must be made by a person, not by AI." OpenAI and Anthropic did not respond to requests for comment.
First-mover advantage
OpenAI's clear lead stems from the earliest generative AI pilot projects in Congress. In 2023, the House Digital Service, an innovation unit under the Office of the Chief Administrative Officer, established an AI working group and issued the first 40 ChatGPT Plus licenses to bipartisan staff. However, Samantha Carter, Chief Communications Officer for the Office of the Chief Administrative Officer, stated that the House's AI procurement strategy "deliberately avoids dependence on a single vendor." The House issued stricter AI usage guidelines in September 2024, limiting the tools that can be used for official business and prohibiting the input of sensitive information. By then, however, ChatGPT had already become the most familiar AI product for the vast majority of offices. "The first company everyone heard about was OpenAI," Harris said. "In the early days of the industry, ChatGPT was almost synonymous with artificial intelligence, much like 'Kleenex' for tissues or 'Band-Aid' for bandages."
OpenAI has also been providing training resources to Congress. The non-partisan Congressional Management Foundation conducts AI training for Congress and has partnered with OpenAI to offer ChatGPT-specific courses for senior staff in both chambers, taught by OpenAI engineers. Several offices, including Fedorchak's, have received product-specific training from OpenAI and Microsoft. Meanwhile, major AI companies are expanding their Washington lobbying teams, closely monitoring AI procurement rules, application standards, and the legislative process. Non-profit watchdog Issue One, analyzing federal lobbying filings, found that OpenAI's federal lobbying spending in the first quarter of 2026 was $1 million, an 82.5% increase year-over-year. Anthropic's lobbying spending was nearly $1.6 million, more than double its spending in the first quarter of 2025. Harris noted that it was only after ChatGPT became established in Congress that lawmakers' teams began to pay attention to competitors. "After the conflict between the Department of Defense and Anthropic, many offices contacted us asking for information about Claude." The background: In January, Anthropic refused to lift restrictions, declining to allow Claude to be used for large-scale domestic surveillance or fully autonomous weapons development. The dispute escalated in early March, with the Department of Defense designating the company as a supply chain risk. No equivalent scale of cooperation between Anthropic and non-partisan congressional training institutions has been found.
Why model selection matters
Eric Petri, a researcher at the Brennan Center for Justice, believes that if government agencies are clearly favoring one company's products, it should be subject to public scrutiny. "It's possible that a particular product is indeed more suitable for government applications or has a technological edge. But if the selection is influenced by partisan preferences or political gamesmanship, the problem is serious." This standard applies whether it's a major government contract bid or a single office subscribing to software. "We should ensure that taxpayer money is used to purchase the best tools on the market, not to favor politically favored companies," Petri said. Harris proposed that before lawmakers begin drafting industry regulations, congressional offices and non-partisan legislative bodies should thoroughly test multiple large language models, comparing their strengths and weaknesses and understanding their operational characteristics. "The future Congress will be responsible for setting AI rules. Lawmaker teams must be fully exposed to a variety of products to understand how different models work."
Comments