WeVote
Log In

Bill

Bill

HB 3720

AI-MEANINGFUL HUMAN REVIEW

104th Regular Session Introduced by Abdelnasser Rashid

Requires continuous meaningful human review and impact assessments for AI systems used by state agencies, with procurement limits and protections for workers and impacted individua

Rule 19(a) / Re-referred to Rules Committee
0
WeVote Research Nonpartisan
Bill Summary · HB 3720

Summary — HB 3720: Meaningful Human Review of Artificial Intelligence Act

Status: Rule 19(a) / Re‑referred to Rules Committee
Introduced: Feb 18, 2025 (Rep. Abdelnasser Rashid) — bill activity includes committee hearings and passage votes in one chamber (see timeline below).

Purpose

To require continuous, substantive human oversight of automated decision‑making systems (ADMS) used by State agencies (or entities acting on their behalf) for functions that affect individuals’ rights, benefits, safety, or welfare — and to require regular impact assessments and transparency to the Governor and General Assembly.

Key provisions

  • Continuous meaningful human review:

    • A State agency (or contractor acting for an agency) may not use an ADMS without continuous meaningful human review when the system:
    • Is used to deliver any public assistance benefit; or
    • Will have a material impact on the rights, civil liberties, safety, or welfare of any individual in the State; or
    • Affects a statutorily or constitutionally provided right of an individual.
    • “Meaningful human review” is defined as review by trained individuals who understand the system’s risks and functionality and have authority to intervene, approve, deny, or modify automated decisions.

  • Procurement limitation:

    • Agencies may not procure, purchase, or acquire services or systems that rely on ADMS for the covered functions unless the system is subject to continuous meaningful human review.

  • Impact assessments:

    • Agencies that use permissible ADMS must conduct an impact assessment signed by the person(s) responsible for meaningful human review.
    • Assessments must be performed at least once every 2 years and after any material change to the ADMS.
    • Required contents (excerpted): objectives of the ADMS; evaluation of the system’s ability to meet those objectives; description/evaluation of system development including a summary of underlying algorithms, computational models and AI tools; testing; and the notification mechanism for impacted individuals (including how individuals are informed of use and their rights/options).
    • Impact assessments must be submitted to the Governor and General Assembly.

  • Employee protections:

    • Use of ADMS shall not result in discharge, displacement, reduction in hours/wages/benefits, transfer of duties away from employees, or impair collective bargaining or civil service status. Existing rights and bargaining relationships must be preserved.

Definitions / scope

  • “Automated decision‑making system” is broadly defined to include software that processes data and applies rules or machine‑learning algorithms to produce conclusions, recommendations, or predictions. The definition excludes basic internal tools (e.g., calculators, spellcheck, office-supply ordering) that do not materially affect rights, liberties, benefits, safety, or welfare.

Who is affected

  • State departments, boards, commissions, public authorities, and contractors/vendors.
  • Individuals receiving public assistance or otherwise subject to decisions that affect rights, safety, welfare, or benefits.
  • State employees — protections aim to prevent job displacement or erosion of collective bargaining rights.

Procedural timeline (selected)

  • Filed: 2025-02-07 / First reading: 2025-02-18
  • Assigned to committees, hearings and substitute considered in April 2025
  • Passed in one chamber and received from the House: actions on 2025-05-14 and 2025-05-15
  • 2025-03-21: Rule 19(a) / Re‑referred to Rules Committee (current status as provided)

Notes / open questions

  • The available text is partially truncated in places; later sections (full testing requirements, details of algorithm disclosure, enforcement mechanisms, sanctions, or exemptions) are not fully shown here.
  • The bill requires agency reporting and human oversight but does not detail criminal or civil penalties in the excerpt provided.
  • Implementation will likely require agency policies, training programs, procurement rule changes, and procedures for producing and securing impact assessments.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.