OpenAI’s ChatGPT has quietly become Capitol Hill’s default AI tool. According to TechCrunch AI, citing a CNBC analysis of House disbursement records, OpenAI took home roughly 90% of all money House offices, committees, and institutional accounts spent on AI tools in the year ending March 31. That’s not a lead. That’s a rout.
The numbers tell the story. Congress spent around $100,580 across 798 transactions on ChatGPT, out of at least $113,740 in total AI spending. Anthropic’s Claude landed a distant second at $13,160 across just 37 transactions. Everyone else barely registered.
What stands out here is how one product locked up a market before most people noticed there was a market at all.
The partisan split nobody’s talking about
Here’s a detail worth sitting with. Democratic offices accounted for $54,165 in AI purchases, more than three times the $15,782 spent by Republican offices, per the CNBC data TechCrunch AI reported.
That gap matters. AI adoption inside government isn’t landing evenly, and spending patterns often forecast who writes the rules later. The offices using these tools daily tend to understand their limits and their leverage. The offices sitting out risk legislating on tools they’ve never actually touched.
One caveat keeps this honest: the figures exclude free accounts and AI bundled into larger software contracts. Real usage on both sides is almost certainly higher. Paid spending is just the part we can see.
What staffers are actually doing with it
This isn’t experimentation for its own sake. Congressional staffers are putting these tools to work on core legislative labor, according to TechCrunch AI. The list includes:
- Writing memos
- Summarizing and analyzing legislation
- Responding to constituents
- Preparing hearing materials
- Sorting through policy research
- Drafting social media posts
Read that list again. Summarizing legislation and analyzing bills is the actual job. When the tool touches the substance of governance, not just the busywork around it, adoption stops being a novelty and starts being infrastructure.
Why this is significant
Government moves slow on new technology. It usually trails the private sector by years. So a near-total ChatGPT sweep inside the House, this fast, signals how deep the default has already set.
This is significant because defaults compound. Once staffers build workflows around one tool, switching costs climb. Training, prompts, and habits harden around the incumbent. OpenAI isn’t just winning transactions here. It’s winning the muscle memory of the people who help write federal law.
There’s also a regulatory wrinkle. The same institution weighing AI oversight is now running on a single vendor’s product. That raises fair questions about dependence, procurement diversity, and whether lawmakers can evaluate an industry they increasingly rely on to do their own jobs.
The takeaways
For AI vendors: the enterprise and government land grab is happening now, not later. OpenAI didn’t win with a splashy government contract. It won with bottom-up adoption, one staffer and one $20 subscription at a time. Anthropic and the rest have ground to make up, and the window narrows as workflows calcify.
For businesses watching their own AI rollout: this is the playbook in miniature. Tools that get into individual hands early become the standard before procurement ever runs a formal bake-off. If you want a different tool to win inside your org, you have to fight the default before it forms.
For everyone tracking AI policy: watch who’s actually using these systems. The offices spending real money are building fluency. When AI legislation comes, and it will, that fluency shapes the outcome more than any hearing soundbite.
ChatGPT owns Congress’s AI budget today. The more interesting question is whether that translates into influence over how the technology gets governed tomorrow. You can find the full breakdown in the original report from TechCrunch AI.