Artificial intelligence tools are increasingly being used by House staffers and outside groups to draft legislation, a development that is raising alarms about the potential for errors and unintended legal consequences. Experts and congressional members warn that AI, while useful for many tasks, is not yet capable of the nuanced legal drafting required for effective laws.
Staffers and advocacy groups are reportedly turning to AI platforms like Claude and ChatGPT for drafting legislative language and reviewing bills. However, the output from these tools often contains inaccuracies and misses critical legal distinctions, such as the difference between a tax credit, deduction, exclusion, or grant. This has led to the Office of Legislative Counsel (OLC), which provides non-partisan legal drafting services to the House, spending more time correcting AI-generated text than it would take to draft legislation from scratch.
Representatives Joe Morelle and Norma Torres have voiced concerns, emphasizing that AI cannot replace the expertise of OLC professionals and that the ultimate responsibility for legislation rests with members of Congress and their staff, not with chatbots. Wade Ballou, former head of the OLC, noted that AI might exclude entities like tribal nations if a bill defines a state narrowly, and Ari Hershowitz, a lawyer who has worked with the OLC, warned that AI could incorrectly cite statutes, introducing significant flaws into the legal system.
Daniel Schuman, executive director of the American Governance Institute, stated that commercially available AI tools lack the necessary attention to detail and legal understanding for bill drafting. Marci Harris, executive director of the POPVOX Foundation, underscored the critical role of the OLC as a quality-control check essential for Congress's constitutional power of lawmaking.