House AI Spending: ChatGPT Accounts for 80%

OpenAI Dominates Early AI Spending in Congress, Shaping Regulatory Landscape

The rapid integration of artificial intelligence within the U.S. Congress is largely being driven by a single dominant player: OpenAI. A recent CNBC analysis of House disbursement records reveals that OpenAI’s flagship product, ChatGPT, accounted for an overwhelming majority of identifiable AI spending by House offices and committees during the year ending March 31, 2026. This significant expenditure underscores not only the burgeoning adoption of AI tools by lawmakers but also the intense competition among leading AI firms to influence the direction of AI regulation in Washington.

The data, which covers spending on named AI vendors between April 1, 2025, and March 31, 2026, shows a total of at least $113,740. Of this, ChatGPT garnered an impressive $100,580 across 798 transactions, representing approximately 88% of the total expenditure and 96% of all recorded transactions. In stark contrast, Anthropic’s Claude, a key competitor in the generative AI space, trailed significantly, with $13,160 spent across 37 transactions.

This spending pattern is more than just a snapshot of office productivity tools. It signals a strategic battleground where major AI developers, including OpenAI, Anthropic, Google, and Microsoft, are vying for influence as Congress and federal agencies grapple with the complex questions of AI procurement, usage, and oversight.

Congressional staff are leveraging these AI tools for a range of critical tasks, from summarizing complex legislation and drafting memos to preparing hearing materials and responding to constituent inquiries. These applications directly impact how offices understand policy, communicate with voters, and manage the immense volume of information flowing through Capitol Hill.

Representative Julie Fedorchak, R-N.D., highlighted the transformative impact of AI on her office’s workflow. Her staff previously dedicated hours to generating briefing memos for her numerous daily meetings. By employing ChatGPT to build a searchable database of her legislative work and constituent materials, and subsequently using AI to generate briefing memos, Fedorchak stated that “dozens and dozens of hours every week of staff time” are now being saved.

Marci Harris, CEO of the POPVOX Foundation, a non-profit dedicated to AI education for congressional staff, confirmed that other offices are similarly utilizing AI for summarizing reports and bills, drafting social media content, processing constituent communications, analyzing legislation, and sifting through policy research. However, Harris also pointed out a paradoxical effect: while AI can streamline tasks, it may also be exacerbating the overwhelming nature of information for already stretched staff, as lobbyists and constituents also increasingly employ AI to deliver lengthier and more frequent communications. This dynamic is occurring against a backdrop of historically reduced staff levels in Congress, as noted by Maya Kornberg, a senior research fellow at the Brennan Center for Justice.

The analysis also revealed a partisan divide in visible AI spending. Democratic member offices accounted for $54,165 in identified AI purchases, more than triple the $15,782 spent by Republican offices. ChatGPT transactions appeared in 44 Democratic offices and 27 Republican offices, while Claude was utilized in five Democratic offices and one Republican office. This spending disparity highlights a nuanced political reality: Democrats are leading in tangible AI adoption, even as many within the party have voiced concerns about the technology’s potential risks to privacy, civil rights, elections, and the concentration of corporate power.

**The Elusive Full Picture of Congressional AI Investment**

It is crucial to acknowledge that the figures derived from the CNBC analysis represent an incomplete picture of Congress’s total AI engagement. The disbursement records primarily capture explicit expenditures on named AI vendors. This methodology excludes the use of free versions of popular AI tools, as well as AI functionalities integrated into broader software contracts, such as Microsoft’s Copilot. Furthermore, some offices may use internal reimbursement systems where specific AI tool purchases are not itemized in public records.

The Senate’s public spending reports also present a challenge, as they do not identify software vendors in the same granular manner as the House, making it difficult to trace specific AI subscriptions, including those for ChatGPT. According to chamber rules, Claude is not permitted for official use within the Senate.

Harris suggested that the House’s deployment of Microsoft Copilot could represent a significant portion of official AI spending that is not explicitly captured in the current analysis of named AI vendor transactions.

Many lawmakers are also implementing internal guidelines to govern AI usage. Representative James Walkinshaw, D-Va., emphasized that while AI assists with research, social media, and newsletters, all factual assertions must be independently verified, and all critical decisions remain human-led. “Our caseworkers in our district office can use AI to help organize casework requests, maybe expedite them, move things along faster,” he stated. “But when they’re making decisions related to a constituent’s case — are we going to advocate for our constituent with the Social Security Administration or with IRS? — that has to be a human decision, not an AI decision.”

Neither OpenAI nor Anthropic provided comment when contacted by CNBC.

**A Strategic Head Start: OpenAI’s Early Influence**

OpenAI’s apparent early lead in congressional AI adoption can be traced back to the chamber’s initial formal explorations of generative AI. In 2023, the House Digital Service initiated an AI working group and distributed the first 40 ChatGPT Plus licenses to a bipartisan group of staff. While Samantha Carter, the Chief Communications Officer for the Chief Administrative Officer’s office, stressed that the House’s approach “is deliberately not limited to a single vendor,” the early introduction of ChatGPT proved pivotal.

Despite more restrictive guidance issued in September 2024, which cautioned against the use of sensitive information and outlined permissible tools, ChatGPT had already become the de facto AI product for many congressional offices. As Harris observed, “OpenAI was the firm that they heard about. In the early days, ChatGPT was almost synonymous with AI — the way Kleenex is with tissues or Band-Aid is with bandages.”

OpenAI has also actively cultivated relationships within Congress through training initiatives. The Congressional Management Foundation, a nonpartisan organization, has partnered with OpenAI on ChatGPT training sessions for senior House and Senate staff, with OpenAI engineers contributing to program delivery. Some offices, including Representative Fedorchak’s, have received personalized training on products from both OpenAI and Microsoft.

Concurrently, AI companies are intensifying their lobbying efforts in Washington. OpenAI reported $1 million in federal lobbying expenditures in the first quarter of 2026, an 82.5% increase from the previous year, according to Issue One, a non-profit government watchdog. Anthropic’s lobbying spending nearly reached $1.6 million, more than tripling its expenditure from the same period in 2025.

Harris noted that interest in OpenAI’s competitors gained momentum only after ChatGPT had established a significant presence within congressional offices. The high-profile dispute involving Anthropic and the Department of Defense, which included Anthropic’s refusal to lift restrictions on Claude for mass domestic surveillance and fully autonomous weapons, and the Pentagon’s subsequent designation of the company as a supply-chain risk, appears to have spurred greater interest in Claude among some congressional staff. Notably, CNBC has not identified a comparable partnership between Anthropic and a non-partisan Hill training group.

**The Significance of AI Model Choice**

Eric Petry of the Brennan Center argues that scrutinizing the government’s apparent preference for one company’s product over another is warranted. He posits that while one product might indeed be superior for specific government applications, a choice driven by “partisan preferences, or there’s politics at play, then it becomes really concerning.” This principle, he asserts, applies equally to awarding large government contracts and to individual office procurement of software subscriptions. “We want to make sure that taxpayer dollars are being used to give the government the best tools available, and not to reward political favorites,” Petry stated.

To effectively regulate the AI industry, Harris contends, congressional offices and non-partisan legislative agencies must have the opportunity to test and compare various AI models, understand their capabilities and limitations, and assess their operational behaviors. “It’s going to be their job to regulate it, so they need to be able to get their hands on all of them and understand the way that they work,” she concluded.