Requests for legislation coming into Congress’ Office of Legislative Counsel were up 72% in 2025 compared to two years earlier. That number came from Warren Burke, who leads the office, in remarks to the House last year. Four anonymous congressional staffers now say they know part of what’s driving it.
The staffers, who apparently help draft bills, spoke to a story published Monday. Their claim: AI-generated text is seeping into bills at an alarming rate.
Alarming for a specific reason. When bills pass, they turn into the federal statutes that are part of the bedrock of American life.
“It’s absolutely terrifying,” one of them said.
The people writing the language often aren’t in the building
One of the anonymous staffers said people and organizations outside Congress are a big part of the issue. They’re “going to Claude or ChatGPT to draft the legislative language itself,” the person said.
That’s the part worth sitting with. Not a staffer cleaning up a paragraph at midnight, but outside groups handing over machine-drafted statutory text and expecting the professionals to sort it out.
And sorting it out is the expensive part. The Office of Legislative Counsel is a collection of mostly lawyers who format and draft laws. According to the same staffer, the office spends “more time trying to fix AI-drafted legislation than it would take for them to draft it from scratch.”
Read that as a productivity story and it inverts. The tool that’s supposed to save time is generating net new work for the one office that can’t refuse it.
We’ve seen this exact failure before, in court filings
If the report is true, AI in the legislative process sounds like it does something similar to AI in the lawsuit-writing process. It gives more people confidence to try and generate documents in jargon-heavy language that needs to be precise. It also makes more work for the people who eventually have to read those documents.
Legislative text is a bad fit for a system that’s fluent by default. Statutes live or die on precision, and text that sounds like a law is not the same thing as text that functions as one.
The Luna incident, and the tweet that disappeared
In a puzzling and somewhat ambiguous incident in June, Representative Anna Paulina Luna, a Republican from Florida, posted apparent evidence online that Anthropic’s Claude chatbot may have been used by a staffer to draft legislation.
Later, she seemed to confirm that it had been used to at least edit a bill.
But that tweet was deleted. It was replaced with a tweet saying the chatbot was only used for a summary of the bill.
Draft, edit, summarize. Three different claims about the same bill, narrowing each time, with the evidence trail getting shorter as it went. That’s the whole problem in miniature: nobody outside the office can tell which one is true, and the disclosure isn’t required.
The staffers are anonymous, which limits how far you can push any of this. But the 72% figure is on the record, from the person who runs the office absorbing the work, and it points the same direction the staffers do.
