While tools like ChatGPT and Claude allow congressional staffers and advocacy groups to generate text quickly, the resulting drafts are often filled with technical errors and incorrect legal citations.
According to officials interviewed by Politico, these flawed submissions frequently take longer for human experts to correct than if the lawyers had written the bills from scratch.
This influx of error-prone text creates significant risks for the American legal infrastructure, as mistakes in statutory language can trigger unintended consequences or costly lawsuits.
Minor errors in an AI-drafted bill, such as misidentifying a tax credit as a grant or using narrow definitions that accidentally exclude tribal nations, could change how federal programs are managed for millions of people.
Because the office acts as a critical quality-control check for Congress, the increased time spent debugging AI-generated text threatens to slow down the entire lawmaking process.
While many staffers continue to use AI for daily work, the legislative counsel is exploring more specialized technology to manage the workload.
The office has integrated tools like the Comparative Print Suite, which uses natural language processing—a branch of AI focused on interpreting text—to help lawyers visualize how new bills would alter the existing U.S.
Code.
Unlike general-purpose chatbots, these specialized systems are designed to flag uncertainties rather than generate "hallucinations," or fabricated information, ensuring that final laws accurately reflect the intent of elected officials.