You might not expect it, but the most popular AI tool on Capitol Hill isn't some government-specific secure AI platform — it's ChatGPT, the same consumer product anyone can sign up for.

According to a full year of U.S. House of Representatives procurement records from April 2025 to March 2026, the House spent at least $113,700 on AI tools over that period. ChatGPT alone accounted for $100,600 across 798 transactions — roughly 88% of total AI spending. The remaining $13,000 was split among other AI tools, none of which came within a rounding error of ChatGPT.

What Are Members of Congress Doing with ChatGPT?

TechCrunch's analysis of House spending records reveals ChatGPT's use on Capitol Hill runs far deeper than people might assume.

First, drafting memos and bill summaries. Members of Congress and their staff need to read massive volumes of legislative text — a single comprehensive bill can run hundreds or thousands of pages. Using ChatGPT to quickly summarize core content and generate memo drafts delivers order-of-magnitude efficiency gains. Final bill language is always human-written and reviewed, but initial drafts and summarization are already heavily delegated to AI.

Second, legislative research. When members need to understand the state of a policy area, comparable legislation in other countries, or relevant academic research, ChatGPT is the fastest starting point — it can produce a structured overview in seconds, and staff then dive deeper into primary sources based on that overview.

Third, constituent communications. Congressional offices receive massive volumes of constituent letters and emails daily that need triaging, summarizing, and response drafting. ChatGPT can quickly identify email topics, categorize them by policy area, and generate draft responses for staff review and sending. The 798 transactions figure itself indicates ChatGPT isn't just one office's experiment — it's widely adopted across the House.

There are several reasons this finding is notable.

First, government AI procurement wasn't won by a "national champion." Many might assume government agencies would use AI platforms built specifically for government with strict security certifications, but Congress chose the consumer-grade ChatGPT. This suggests that in real-world usage, security certifications and customization may not be prioritized as highly as we think — being good enough, accessible, and available right now matters more than anything.

Second, AI is penetrating the core processes of democracy. Legislation is one of democracy's most core functions. When AI starts participating in bill drafting, policy research, and constituent communications — regardless of depth — it means AI is no longer merely an "efficiency tool": it's shaping, to some degree, the language of laws, the framing of policies, and how government communicates with citizens.

Security and Accuracy Concerns

Of course, this state of affairs raises more than a few concerns.

The most immediate issue is data security. Members of Congress handle a lot of sensitive information — involving national security, non-public legislative deliberations, constituents' private information. Inputting that information into ChatGPT — does that comply with government data security rules? OpenAI offers ChatGPT Enterprise and government editions, but it's unclear from procurement records whether the House is purchasing regular ChatGPT subscriptions or enterprise/government versions. If it's the regular version, where data may be used for model training, the risks are significant.

The second issue is AI hallucination and accuracy. ChatGPT fabricates cases that don't exist, cites statutes that aren't there, produces incorrect policy analysis. If congressional staff cite AI-generated content without verification, it could lead to flawed legislative reasoning or inaccurate responses to constituents. Professional congressional staff will of course verify, but under time pressure, the "looks right" hallucination effect can lower vigilance.

The third issue runs deeper: inequality in AI usage. ChatGPT's proliferation on Capitol Hill means offices that use AI effectively are vastly more efficient than those that don't. Could this skew legislative resources toward AI fluency? Could smaller offices without the budget for AI subscriptions or staff skilled in AI tools find themselves at a legislative disadvantage?

Like it or not, AI is already on Capitol Hill. It's not a vendor waiting at the door for approval — it's already on members' desks, in bill drafts, in responses to constituents. That 88% figure tells us: in AI tool competition, the winner isn't the most secure, isn't the most customized, isn't the cheapest — it's the most usable.

When the people writing laws are using ChatGPT to write them, the debate over AI regulation takes on an extra layer of irony.

See you tomorrow.