Congressional Staff Now Draft Legislation With AI Chatbots

Congressional offices have begun integrating generative AI tools into the legislative process, with thousands of House and Senate staffers now using chatbots to draft bills, amendments, and constituent communications. The shift promises efficiency but introduces new risks into an already error-prone system.

Widespread adoption across Capitol Hill

According to reporting first published by The Washington Post, both chambers have authorized staff to use major AI platforms including Microsoft Copilot, ChatGPT, Gemini, and Claude for official work. The House purchased 6,000 Copilot licenses last year, with roughly half currently in active use.

Broad guidelines prohibit staffers from entering sensitive constituent information into chatbots, generating deepfakes, making personnel decisions, or finalizing legislation through AI. Political uses—such as criticizing opposition parties or specific politicians—are also banned. Beyond these baseline rules, individual congressional offices are encouraged to develop their own AI policies, creating a patchwork of hundreds of different standards across Capitol Hill.

Review offices overwhelmed by AI-generated text

The House Office of Legislative Counsel, responsible for catching errors in proposed legislation, has been strained by the influx of AI-drafted content. As Politico reported, the office's lawyers now spend significantly more time reviewing and rewriting AI-generated proposals compared to human-drafted text.

Warren Burke, who leads the Office of Legislative Counsel, testified in March that material "produced by outside entities, whether a human or a computer, often takes significantly longer to review and revise as necessary to ensure the text accomplishes the sponsor's intended policy goals."

Legislative text is notoriously complex, frequently citing previously passed laws and amending specific sections of the U.S. Code. Even minor errors can create major problems. A single typo in a 2006 spending cuts bill caused the House and Senate to pass slightly different versions, triggering a potential constitutional crisis. This year, the House had to recall a bill from the Senate after discovering it was missing an amendment that had passed on the House floor.

Why it matters

The integration of AI into legislative drafting represents more than a workflow change—it threatens to degrade the institutional knowledge required to craft precise, enforceable laws. Every word and punctuation mark in federal legislation carries legal weight. As staffers increasingly outsource drafting to chatbots, the risk grows that lawmakers will lose the deep understanding needed to anticipate how their words will affect millions of Americans. The technology may also increase the frequency of legislative errors at a time when the review process is already strained, potentially introducing flaws into the U.S. Code that could prove costly to fix.

Loss of legislative expertise

Lawmakers already rely heavily on staff to translate policy ideas into legislative language. The additional layer of AI mediation raises questions about whether anyone involved fully grasps the implications of the text being written into law. The concern extends beyond simple accuracy to the erosion of specialized skills that have traditionally ensured legislative precision.

These details were first reported by The Washington Post and Politico.