TallyIDAHOLegislative Tracker
H06872026 Regular Session

Adds to existing law to establish provisions regarding unbiased artificial intelligence in state government purchasing.

PROCUREMENT -- Adds to existing law to establish provisions regarding unbiased artificial intelligence in state government purchasing.

IntroducedIn CommitteeFloor VoteEnacted

Committee: State Affairs

▶ Show statement of purpose

The purpose of this legislation is to ensure that artificial intelligence systems used by state government operate as neutral, accurate, and reliable tools that serve the public interest. As agencies increasingly rely on large language models to assist with research, analysis, and decision-making, this act establishes standards to prevent the intentional distortion, suppression, or prioritization of information to advance undisclosed ideological agendas, including diversity, equity, and inclusion. The legislation affirms that artificial intelligence may describe or analyze any lawful subject in an informational or academic manner, while requiring transparency, procurement safeguards, and oversight to protect objectivity, public trust, and the integrity of state government operations.

▶ Show fiscal note

This legislation causes no increase in revenue, or additional expenditure of funds at the state or local level of government, therefore, this legislation has no fiscal impact.

▶ Show full bill text

LEGISLATURE OF THE STATE OF IDAHO Sixty-eighth Legislature Second Regular Session - 2026 IN THE HOUSE OF REPRESENTATIVES HOUSE BILL NO. 687 BY JUDICIARY, RULES AND ADMINISTRATION COMMITTEE AN ACT1 RELATING TO PURCHASING; AMENDING CHAPTER 92, TITLE 67, IDAHO CODE, BY THE2 ADDITION OF A NEW SECTION 67-9236, IDAHO CODE, TO ESTABLISH PROVI-3 SIONS REGARDING THE PURCHASE OF UNBIASED ARTIFICIAL INTELLIGENCE IN4 STATE GOVERNMENT, TO DEFINE TERMS, TO ESTABLISH REQUIREMENTS FOR THE5 PURCHASE, DEPLOYMENT, OR USE OF A LARGE LANGUAGE MODEL, TO ESTABLISH6 CONTRACT REQUIREMENTS, TO PROVIDE FOR COMPLIANCE REVIEW AND PROCE-7 DURES, AND TO PROVIDE FOR CONSTRUCTION; AND DECLARING AN EMERGENCY.8

Be It Enacted by the Legislature of the State of Idaho:9

SECTION 1. That Chapter 92, Title 67, Idaho Code, be, and the same is10 hereby amended by the addition thereto of a NEW SECTION, to be known and des-11 ignated as Section 67-9236, Idaho Code, and to read as follows:12 67-9236. UNBIASED ARTIFICIAL INTELLIGENCE IN STATE GOVERNMENT. (1) As13 used in this section:14 (a) "Agency" means any department, office, division, board, commis-15 sion, institution, or other entity of the executive branch of state gov-16 ernment, including institutions of higher education.17 (b) "Agency head" means the chief executive officer or governing au-18 thority of an agency.19 (c) "Diversity, equity, and inclusion" or "DEI" means any trainings,20 programs, activities, or instruction that is derived from or that pro-21 motes the tenets or concepts of critical theory, including but not lim-22 ited to the concepts of unconscious or implicit bias, microaggressions,23 internalized racism, cultural appropriation, structural equity, set-24 tler colonialism, group marginalization, systemic oppression, social25 justice, institutional or systemic racism, white fragility, racial26 privilege, disparate impact, intersectionality, sexual privilege,27 patriarchy, gender theory, queer theory, neopronouns, transgender28 ideology, misgendering, othering, deadnaming, heteronormativity,29 allyship, or any other related formulation of these tenets or concepts.30 (d) "Ideological agenda" means the intentional alteration, suppres-31 sion, prioritization, or framing of information or representation in32 an artificial intelligence system's outputs to promote a political, so-33 cial, or policy outcome, including DEI concepts, when such alteration34 is not expressly requested by the user.35 (e) "Large language model" means a generative artificial intelligence36 system trained on large datasets that is capable of producing natural37 language text, images, or other content in response to user prompts.38 (2) An agency shall procure, deploy, or use a large language model only39 if the model is developed and implemented in accordance with the following40 principles:41

2 (a) Truth-seeking: A large language model shall prioritize factual ac-1 curacy, historical fidelity, scientific inquiry, and objectivity when2 responding to prompts seeking information or analysis. Where reliable3 information is incomplete, disputed, or unavailable, the model shall4 acknowledge such uncertainty; and5 (b) Ideological neutrality: A large language model shall function as6 a neutral, nonpartisan tool and shall not intentionally manipulate out-7 puts to advance or suppress an ideological agenda, including DEI con-8 cepts. Ideological, normative, or policy-based responses may be gener-9 ated only when expressly requested by the user or when clearly disclosed10 and readily accessible to the user.11 (3)(a) No agency shall procure, deploy, or use a large language model12 that is designed, trained, fine-tuned, or configured to:13 (i) Alter factual information or historical representation for14 the purpose of achieving outcomes associated with DEI;15 (ii) Suppress, refuse, or condition the generation of lawful con-16 tent based on the race or sex of individuals or groups;17 (iii) Embed DEI-based constraints, guardrails, or response poli-18 cies that override accuracy or objectivity; or19 (iv) Require users to adopt, affirm, or comply with DEI concepts20 as a condition of receiving information.21 (b) Nothing in this subsection shall prohibit a large language model22 from accurately describing DEI concepts, history, or critiques in an23 informational, academic, or analytical context when such information24 is requested by the user.25 (4)(a) Any contract entered into by an agency for the procurement or use26 of a large language model shall require compliance with this paragraph.27 (i) A contract shall require the vendor to provide the agency with28 reasonable access, subject to confidentiality and nondisclosure29 protections, to materials necessary to verify compliance with30 this section, including but not limited to:31 1. System prompts and content policies;32 2. Model specifications and training directives;33 3. Evaluation methodologies and testing results; and34 4. Source code, configuration files, or other backend tech-35 nical materials, if necessary to assess intentional ideo-36 logical manipulation.37 (ii) A contract shall provide that, upon a finding of material38 noncompliance following notice and a reasonable opportunity to39 cure, the agency may terminate the contract and require the vendor40 to bear reasonable decommissioning or transition costs.41 (b) Agencies may utilize state personnel or independent technical ex-42 perts to conduct compliance reviews.43 (5) Each agency shall adopt procedures to ensure compliance with this44 section, including periodic review of large language model performance. In45 evaluating compliance, agencies shall consider:46 (a) Whether distortions or biases are systemic or intentional rather47 than incidental errors;48 (b) The intended use and operational context of the model; and49 (c) The technical feasibility of corrective action.50

3 (6) Nothing in this section shall be construed to:1 (a) Regulate private speech or restrict the development or sale of ar-2 tificial intelligence systems outside state government;3 (b) Mandate ideological viewpoints or prohibit lawful academic in-4 quiry; or5 (c) Create a private right of action.6

SECTION 2. An emergency existing therefor, which emergency is hereby7 declared to exist, this act shall be in full force and effect on and after its8 passage and approval.9

house Chamber· Mar 2, 2026

House Third Reading

✓ Passed
56 Yea
11 Nay
3 absentPassed by 45 votes
Republican
56 yea/3 nay
Democrat
0 yea/8 nay
Show all 67 voter names

Introduced, read first time; referred to: State Affairs