AI-Generated Legislation in the US Congress
Members of Congress are increasingly using tools like ChatGPT and Claude to draft legislation. While these systems offer convenience, attorneys in the House of Representatives now find themselves spending more time reviewing AI-produced documents than they would writing new bills from the ground up. The reason is a steady stream of legal errors in the drafts. According to Politico, congressional insiders say lawyers often encounter problems with legal terminology, references to existing statutes, and the scope of the word “state.” For instance, an AI model might be unable to tell the difference between:
- a tax credit;
- a tax deduction;
- a tax exemption;
- a grant.
These drafts also tend to cite established laws incorrectly.
How AI-Produced Bills Are Checked and Edited
The House Office of the Legislative Counsel (OLC) has the job of reviewing and revising AI-assisted legislation. Wade Ballou, who led the OLC for nearly a decade until his departure in 2024, described the challenge:
“Artificial intelligence does not always understand the important legal nuances that are critical in legislative practice.” — Wade Ballou
One recurring issue, he noted, is that AI may fail to include Washington, D.C., or Native American tribal nations in relevant sections of a bill.
Another risk is that congressional staffers who rely on AI for drafting may themselves have a limited understanding of the resulting text. In response, the OLC has built a custom internal tool called Comparative Print Suite. It allows staff to see how a proposed bill would change current law. Unlike publicly available AI services, Comparative Print Suite refuses to invent amendment locations and instead reports an error when it cannot determine where a change should be placed.
The rise of AI in legislative work is part of a larger global movement toward automation and digital transformation in many sectors. However, the ongoing need to correct AI-generated legislation shows that the technology is still far from reliable without expert human intervention. For congressional lawyers, every AI-produced bill is a reminder of the critical role they play in preserving legal accuracy and coherence. The experience offers a cautionary example for other governments and institutions weighing the adoption of similar tools.
As the use of AI in legislative drafting continues to evolve, the implications of such technology extend beyond Congress. Recently, in a groundbreaking case, a US court imposed sanctions for undisclosed AI directives in legal documents, highlighting the growing scrutiny surrounding AI-generated content. This development raises important questions about the reliability and transparency of AI tools in various fields, including law. For more on this significant case, see the recent court ruling on hidden AI instructions.