AI tools such as ChatGPT and Claude are increasingly being used by congressional staffers and outside groups to draft legislation. The problem is that the House Office of Legislative Counsel, which reviews and prepares legislative language, is finding that AI-generated bills can take longer to correct than drafting them from scratch. Officials say the growing use of AI is creating a new workload problem for an office that was already handling a very large volume of legislative requests.
The difficulty comes from the fact that small wording differences can have major legal consequences. AI may struggle to distinguish between concepts such as a tax credit, deduction, exclusion, or grant, while apparently simple definitions can accidentally exclude groups or jurisdictions. Incorrect statutory citations are another concern because they can introduce errors directly into the legal framework. The article quotes legal experts who argue that AI is useful for many tasks but is not yet reliable enough to independently produce legislation ready for enactment.
The scale of the workload is significant. The Office of Legislative Counsel prepared 30,494 bills during the 118th Congress, compared with 10,564 that were actually introduced. It also prepared more than 21,000 amendments. The office had 78 attorneys and 23 support staff in the first session of the current Congress, while its legislative counsel has warned that AI-generated material could increase rather than reduce the workload because externally produced text often requires extensive review and revision.
The irony is that the House is also experimenting with AI internally. Its staff is evaluating Microsoft Copilot for tasks such as legal research, while the office already uses a natural-language-processing system called the Comparative Print Suite to identify how proposed legislation would alter the U.S. Code. The bigger lesson is that AI may make it dramatically easier to produce legislation, but that does not necessarily make legislation easier to validate, interpret, or make legally precise. As AI-generated text floods existing institutional systems, the bottleneck may shift from writing to verification and accountability.
