The Office of Legislative Counsel (OLC) of the U.S. House of Representatives has faced a new burden: the flow of draft bills created with the help of generative AI has noticeably increased. Current and former employees of the agency I interviewed confirm that the problem has already gone beyond isolated cases.
The crux of the difficulty lies in the review stage. Algorithms generate texts that look convincing but contain broken references to current regulations, erroneous definitions, and wording that does not match the legislator's intent. As a result, lawyers have to spend more time correcting such material than writing a document from scratch. Any inaccuracy risks delaying the adoption of an initiative or provoking litigation.
Congresswoman Norma Torres warned of a growing danger: excessive reliance on AI increases the likelihood of errors and wording that distort the legislator's will.
"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 office's head, Warren Burke, attributed the increase in requests to overall legislative activity and the flow of amendments. However, as early as 2024, POPVOX experts noted that large language models simplified the creation of pseudo-bills, which directly increased the volume of work for lawyers.
Congress is not giving up on AI
Despite the problems with drafts, the House of Representatives continues to integrate the technology into its processes. According to Aubrey Wilson, Director of Government Innovation at the POPVOX Foundation, work with Microsoft Copilot in the lower chamber continued in 2026, and since February, employees have been trained to use the tool. Wilson insists that AI should be an auxiliary tool that enhances human work, not a replacement for it.
The OLC itself is also exploring the capabilities of neural networks. In April, Burke reported that the office, together with other legislative branch structures, is analyzing AI application scenarios, including the associated risks.
Significantly, the problem is not limited to the United States. Recently, Guardian journalists found traces of AI hallucinations in a report for the Australian government — the document contained references to non-existent scientific publications.
My analysis: The introduction of AI into lawmaking is a double-edged sword. On the one hand, the technology accelerates routine processes; on the other, it creates an illusion of competence that comes at a high cost during the legal review stage. Until algorithms learn to distinguish real legal norms from plausible fictions, responsibility for the quality of laws must remain exclusively with people. Otherwise, we risk getting a legislative framework built on hallucinations.