ChatGPT Corners 90% Of Capitol Hill's Paid AI Spending

The News

OpenAI's ChatGPT has become the default paid artificial-intelligence tool inside the United States Congress, according to House disbursement records reviewed by TechCrunch. Congressional offices spent around $100,580 on ChatGPT across 798 separate transactions in the year ending 31 March 2026, out of a minimum of $113,740 in total identifiable AI spending.

That works out to roughly 90% of all disclosed House AI outlay flowing to a single vendor. Anthropic's Claude, the nearest rival captured in the records, drew just $13,160 across 37 transactions. No other model developer registered a meaningful share.

The partisan split is striking too. Democratic offices accounted for about $54,165 of the spend, while Republican offices reported $15,782, meaning Democratic-led offices bought roughly three times as much paid AI capacity as their counterparts. Staffers are using the tools to draft memos, summarise legislation, prepare hearing materials, answer constituent mail and write social posts.

The figures almost certainly understate real usage, because they exclude free accounts and any AI folded into larger software contracts.

Why It Matters

Procurement data is one of the few honest signals in the AI market. Marketing decks and download counts can be gamed; line items in a public ledger cannot. When the single largest deliberative body in the world quietly routes nine of every ten AI dollars to one supplier, it tells you which product staff actually reach for under deadline pressure.

That concentration mirrors an earlier software era. The last time a productivity tool captured a public institution this completely was the spread of Microsoft Office through government desktops in the late 1990s, a lock-in that took two decades and a wave of open standards to loosen. Default status compounds: the tool everyone already knows becomes the tool nobody has to justify buying.

The risk for buyers is that habit hardens into dependence long before anyone runs a serious competitive evaluation. A $100,000 annual line item is trivial against a legislative budget, which is precisely why it escapes scrutiny. Small, unexamined purchases are how single vendors end up owning a workflow.

Indian Angle

India is writing its own version of this ledger right now, and it has a choice the US Congress never really faced. The IndiaAI Mission, backed by MeitY with a corpus exceeding Rs 10,000 crore, is explicitly meant to seed home-grown models. Startups such as Sarvam AI and Ola-backed Krutrim are building systems tuned for Indian languages and contexts, and Sarvam has been named among the providers for a sovereign foundation model under that mission.

Government procurement in India runs largely through GeM, which gives the state a lever Washington lacks: the ability to weight tenders toward domestic and data-sovereign options before a foreign default sets in. If Indian ministries, PSUs and parliamentary secretariats drift toward ChatGPT purely out of familiarity, the same 90% concentration could quietly repeat here, only with citizen and legislative data leaving Indian jurisdiction.

There is also a hard-currency angle. Every ChatGPT or Claude seat is billed in dollars, so rupee depreciation makes foreign AI structurally more expensive each year, while rupee-priced Indian models keep costs onshore. For CFOs at Indian enterprises and for the government alike, the Capitol Hill data is less a curiosity than a warning about what happens when procurement follows habit rather than policy.

FAQ

How much did Congress spend on ChatGPT?

House disbursement records show roughly $100,580 spent on ChatGPT across 798 transactions in the year ending 31 March 2026. That is about 90% of the minimum $113,740 in total identifiable AI spending, with Anthropic's Claude a distant second at $13,160.

Why did Democratic offices spend more?

The records show Democratic offices spent about $54,165 against Republican offices' $15,782, roughly three times as much. The data does not explain the gap, but it reflects disclosed paid subscriptions only, not free accounts or AI bundled inside larger software contracts.

Does this affect Indian government AI plans?

Indirectly, yes. It illustrates how quickly one foreign vendor can become a default. India's IndiaAI Mission and GeM procurement give policymakers a chance to weight tenders toward domestic, data-sovereign models such as Sarvam and Krutrim before a similar concentration takes hold.

Where can I read the original report?

The underlying analysis of House spending records was reported by TechCrunch, linked in full at the end of this article.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.