Ohio HB 1032 state-agency and employment AI proposal (2026)
AI-generated text. This page was generated using artificial intelligence.
Ohio House Bill 1032
United States — Ohio
Legislative proposal
H.B. 1032, 136th General Assembly; As Introduced[1]
Introduced in the House October 22, 2026 and referred to House Technology and Innovation November 10, 2026[2]
Not enacted; proposed inventory deadlines begin December 31, 2027 and February 1, 2028[1]
November 17, 2026 — official status and current text version; full introduced text reviewed November 3
Ohio House Bill 1032 (136th General Assembly) is a proposal concerning state-agency AI use and automated employment decisions. Representatives Tristan Rader and Christine Cockley are its primary sponsors. The bill was introduced in the House on October 22, 2026 and referred to the House Technology and Innovation Committee on November 10. The latest available text is the 51-page As Introduced version. It is not enacted law.[2][1]
State-agency use and procurement
Proposed R.C. 125.17 would require annual agency AI-system inventories beginning December 31, 2027, a public master inventory beginning February 1, 2028, and ongoing assessments beginning that February to check for unlawful discrimination and disparate impact. The inventory would include system capabilities, decision-making uses, audits, assessment dates, access to personally identifiable information, savings and risks. Its AI definition excludes listed ordinary software and similar technologies.[1]
Proposed R.C. 125.50 would prohibit state agencies and entities acting for them from using AI for functions related to delivering public-assistance benefits. It would also prohibit use in functions materially affecting rights, civil liberties, safety or welfare unless specifically authorized by law, and restrict agency AI procurement to uses specifically authorized by law. For authorized procurement, it would require an independent predeployment bias audit, submission to the administrative-services director and publication of the audit results at least 60 days before deployment, with personally identifiable information redactions permitted.[1]
Employment: distinct scopes
The bill uses different employer definitions for different provisions. Proposed R.C. 4112.025 would operate within the civil-rights chapter, whose employer definition includes the state, political subdivisions and persons employing four or more people in Ohio, and their agents. It would make specified discriminatory adverse employment actions through an automated process unlawful, and require advance written notice of the process, its trade name and the types and sources of personal information it would analyze. Related amendments would require the civil-rights commission and courts to consider evidence, or its absence, of antibias testing and other proactive efforts.[1]
By contrast, proposed Chapter 4145 defines an employer as the state and specified state or local public bodies and their agencies or instrumentalities. Its detailed duties should not be read as applying to every private employer. The chapter's process definition includes tools that make or assist employment decisions, including targeted recruitment, résumé screening, online-interview analysis and third-party-data analysis; ordinary software is excluded insofar as it does not make an employment-related decision.[1]
Public-employer duties and developer responsibilities
For covered processes, Chapter 4145 would require interaction disclosures, notices before collecting personal data, an opt-out from profiling for automated decisions with legal or similarly significant effects, and advance decision notices. Applicants and employees could request reevaluation. Adverse-decision explanations would identify the principal reason, the process's contribution and the data and sources used, with opportunities to inspect and correct data and appeal through human review.[1]
Human reviewers would need decision-changing authority, relevant training and sufficient time to examine outputs and underlying data; they could not rely solely on the automated output. Employers would need intervention procedures and review logs, and could not make an employment decision through the process without human review. Independent bias audits would be required before deployment and annually thereafter, with reports filed and plain-language summaries published within 30 days of completion. The text also provides for auditor approval, five-year record retention and corrective action approved by the commerce director before deploying a process whose latest audit identified disparate impact.[1]
Developers would have to supply information needed for specified employer disclosures and could contractually assume specified duties without an enforceable liability waiver. The chapter also contains retaliation protections, qualified treatment of trade secrets and legally protected information, and notice to an employee organization before testing, deployment or material modification.[1]
Enforcement
Proposed R.C. 4145.11 would authorize attorney-general declaratory or injunctive actions for specified disclosure, review and audit violations, subject to a two-year limit. Its maximum $5,000-per-day civil penalty concerns violation of a resulting court restraining order or injunction after notice; it is not a flat daily penalty for every underlying AI violation. Proposed R.C. 4145.12 separately provides an aggrieved applicant or employee, or an employee organization acting for them, a civil action for specified violations, including damages, equitable relief, costs and reasonable attorney fees, with joint and several employer/developer liability.[1]
Bargaining and displacement reporting
Amendments to R.C. 4117.08 would expressly include public-employer AI use in collective bargaining over wages, hours and employment conditions. Other amendments would add AI or technological-change information to specified plant-closing or mass-layoff notices, and AI-displacement questions or information to unemployment-benefit applications and certain layoff reports. These reporting provisions do not themselves establish a new AI-displacement benefit.[1]