The Digital Draftsmen: How AI-Generated Bills are Rewriting Congressional Lawmaking

The House Office of Legislative Counsel is grappling with an influx of bills written by artificial intelligence tools like ChatGPT and Claude, which are riddled with errors and omissions that could have far-reaching consequences.

One of the most significant issues is the lack of nuance in these generated proposals. AI tools tend to rely on broad definitions and generalities, leading to oversights that might seem minor at first but could have a profound impact on how laws are interpreted and enacted. For example, former head of the OLC Wade Ballou pointed out that AI may struggle to determine whether a particular funding mechanism should be classified as a tax credit, deduction, exclusion, or grant – a distinction with significant implications for policy.

The accuracy of these bills is also a concern. But it’s not just about getting the numbers right; the use of AI-generated proposals contributes to a lack of understanding among congressional staffers about what their bills are intended to accomplish. Staffers who rely on AI to draft their proposals may not be forced to delve deeply into the issues at hand – and as a result, may not have a thorough grasp of the policy’s implications.

This raises fundamental questions about the role of technology in lawmaking. As we increasingly rely on algorithms to generate text, are we sacrificing something essential to the legislative process? The idea that AI can improve efficiencies by streamlining the drafting process is attractive – but it’s worth considering whether this comes at the cost of nuance and understanding.

The OLC has responded by exploring the use of its own AI tool, the Comparative Print Suite. This program helps staffers visualize how a proposal will change current laws – and returns an error if it can’t determine where changes should be made. However, it’s unclear whether these measures will be enough to address the underlying issues.

The use of AI-generated bills in Congress has parallels with debates about AI in journalism and publishing, where concerns have been raised about “hallucinations” making their way into print. In both cases, we’re grappling with the implications of relying on technology to generate content that has real-world consequences.

As AI-generated bills continue to flow through Congress, it’s worth considering whether we’re creating a system where legislation is written by machines rather than humans. This might seem like a distant prospect – but as we rely more and more on technology to draft our laws, we risk losing something essential to the process: human judgment and oversight.

The use of AI-generated bills in Congress deserves closer scrutiny. As we navigate this new landscape, it’s worth asking whether we’re sacrificing nuance and understanding on the altar of efficiency.