The North Dakota Legislative Council (NDLC), the non-partisan legislative support agency based in Bismarck, is gearing up for a technology leap. The Council is training an artificial intelligence system designed to automatically draft summaries of bills and amendments — work that currently takes its legal staff around 100 hours per session to complete.
This move comes as the NDLC targets a rollout of the tool for the 2027 legislative session. Instead of using broad internet-wide data sets, the system is being trained on past bill summaries from North Dakota’s 2019, 2021 and 2023 sessions.
Information Technology Director Cody Malloy explained the pilot effort produces three versions of each title summary and will undergo comparison with human-drafted summaries for accuracy, before full deployment.
For Grand Forks-area and statewide readers, this signals that state government in Bismarck is embracing technology to boost efficiency. Faster bill summaries could mean legislators and the public get quicker access to digestible legislative content. But it also raises questions of accuracy, transparency and the preservation of human oversight in legislative drafting support.
The NDLC currently drafts summaries for roughly 600 bills each session — work that comes after the session ends. According to Legal Division Director Emily Thompson, simply reviewing staff-prepared summaries takes “easily” 100 hours.
The new AI system is built on Meta’s open-source LLaMA 3.2 1B Instruct model and uses the training program Unsloth.
Unlike generic AI services such as ChatGPT that use wide internet data, the NDLC’s model is custom-trained on North Dakota-specific legislative data to better mimic the style and quality the agency requires.
Malloy emphasized that “you always have a human in the loop… ensuring you aren’t just wholesale shooting out documents from AI.”
House Majority Leader Mike Lefor noted that with newer term-limits, institutional knowledge is shrinking, so a tool that can quickly provide accurate bill/ amendment summaries could be especially helpful.
“That could save a huge amount of time if we can really get this trained to where it’s consistently producing accurate results that can be quickly checked and verified.” — Emily Thompson, Legal Division Director, NDLC.
“What almost every state is doing is making sure there is a human in the loop … We’ve insured that — nothing is just automatically generated or published from AI.” — Cody Malloy, IT Director, NDLC.
How well the AI summaries match human-drafted ones and how errors are detected and managed. Will the public have access to the AI-generated drafts, or visibility into how the tool is used?
While the tool aims to free staff time for deeper analysis, how will roles and responsibilities shift?
If summarization becomes faster, this could open access to clearer legislative information for citizens in Grand Forks, Bismarck and beyond.
How will the NDLC monitor and update the system, ensure data privacy, and adapt as legislation and language evolve?
As the North Dakota Legislature prepares for its 2027 session, the NDLC’s move to deploy AI-driven bill summaries marks a turning point in how state legislative support functions work. For voters and stakeholders in Grand Forks and across the state, it offers the promise of faster, clearer access to legislative information — but also carries a caution: the smarter the tool becomes, the more important human judgement and oversight will be. The efficiency gains are real, but they must be coupled with transparency, accuracy and accountability if the public is to fully trust this new era of tech-augmented lawmaking.