Committee requires state agencies to adopt AI guardrails, keeps humans accountable

The Senate State Agencies & Governmental Affairs Committee passed House Bill 1958 on a voice vote after debate that centered on how to allow agencies to use artificial intelligence while guarding against automated decision-making.

Representative Alex Meeks, sponsor of the bill, told the committee the measure would add AI to existing technology-use policies and require agencies to train employees on when and how AI tools should be used. "Some of the concerns related to this, if a state employee were to take proprietary or citizen data and run it through an AI ... that information then potentially becomes the property of that AI engine," Meeks said. He added the bill requires that "a human is always responsible for making a final decision."

Committee members pressed for specifics about scope and enforcement. Senator Dawson said she wants to avoid putting decision-making “at the mercy of a computer,” describing examples where AI summaries had “hallucinated” or missed important points. Meeks replied the bill would permit automated recommendations but preserve a human review step, and that appeals would remain available for human decisions.

Senator King questioned whether agency directors could already set similar policies without statute; Meeks said the statute creates uniform, statutory guardrails and directs agency directors and the Office of State Technology to incorporate AI into employee policies and training.

The committee adopted an amendment that added a Senate co-sponsor and then voted the bill out of committee. There was no public testimony for or against the measure at the committee hearing.

The bill as passed by the committee requires agencies to expand their technology-use policies to address AI, to provide employee training, and to ensure human accountability for final determinations when AI tools are used.

Representative Meeks closed by asking for a favorable vote; the committee granted the request."