The Office of Legislative Counsel (OLC) of the U.S. House of Representatives has faced a new headache: the flow of draft bills created with the help of artificial intelligence has noticeably increased. And this is placing a serious burden on the team responsible for the legal integrity of legislative initiatives.
The problem lies in the review stage. Documents generated by neural networks are plagued by false references to existing regulations, vague definitions, and wording that distorts the author's original intent. As a result, lawyers have to spend more time fixing such "creativity" than writing text from scratch. Every mistake is a potential delay in the adoption of an initiative or, worse, grounds for litigation.
Congresswoman Norma Torres is raising the alarm: excessive trust in generative models increases the risk of legally incorrect texts that do not reflect the legislator's real intentions.
"This responsibility cannot be shifted onto a chatbot," she emphasized.
Official statistics do not allow for an accurate assessment of the scale of AI's impact on the OLC's workload. In April 2025, the head of the service, Warren Burke, attributed the increase in requests to overall legislative activity and an abundance of amendments. However, POPVOX experts noted back in 2024 that large language models simplified the generation of pseudo-bills, increasing the flow of materials for review.
Congress is not giving up on AI
Despite these difficulties, the House of Representatives continues to integrate the technology into its workflows. In 2026, the use of Microsoft Copilot in the lower chamber continued, and starting in February, employees began training on this tool. Aubrey Wilson, Director of Government Innovation at POPVOX Foundation, insists: AI should be an auxiliary tool that enhances human work, not a replacement for it.
The OLC itself is also exploring the potential of neural networks. In April, Burke reported that the service, along with other legislative branch structures, is analyzing AI use cases, assessing both the benefits and the risks.
This situation is a vivid example of how AI is breaking into areas where precision is critical. For now, the technology is not capable of replacing legal expertise, and attempts to save time could result in much greater costs in the future. Interestingly, similar problems have already surfaced in Australia, where traces of AI hallucinations with references to non-existent publications were found in a report for the government. It is obvious that without strict oversight, AI in the legal sphere is a ticking time bomb.