Governments are using AI to draft legislation—what could go wrong?

Governments in the UK, Italy, Brazil, the US, and other countries are already using AI in legislative processes, for tasks such as categorizing public consultation responses, clustering amendments, and drafting rule texts. The UK government says its Consult tool categorized over 50,000 consultation responses into themes in about two hours at a cost of £240, followed by 22 hours of expert manual review. The Italian Senate uses AI to cluster similar amendments and flag potential delaying tactics, while the Chamber of Deputies advances GENAI4LEX-B. The Brazilian Chamber of Deputies expanded the Ulysses project and allows staff to use Claude, Gemini, and GPT. The US federal government plans to use Gemini to accelerate deregulation. Researchers warn that AI used to screen public comments becomes part of the decision-making system, and foreign governments could flood genuine participation with mass-generated comments, creating a legislative-level DDoS. Ruth Fox of the Hansard Society emphasizes the need for manual review and disclosure of AI use, while Philip Wallach of AEI cautions that AI-drafted rules could be overturned by courts on procedural grounds.