Skip to content
LegisGateAI Regulatory Platform
AI Obligation Atlas

Human oversight of automated decisions

Requirements to preserve meaningful human review, intervention, or override of automated and AI-driven decisions.

17 jurisdictions19 laws & guidance12 binding← All obligations

US States7 laws

C.R.S. § 6-1-1701 et seq.

State of Colorado

Binding

Developers and deployers of Automated Decision-Making Technology making consequential decisions about Colorado consumers.

Law · Pending

California SB 947 (Workplace ADS)

State of California

Binding

An employer, as § 1520(d)(3) defines the term and including a state or local governmental entity, that uses an automated decision system (§ 1520(c)). For §§ 1522(b) to (e) and 1524, an employer that makes a disciplinary or termination decision and, as the section provides, relies solely on an ADS (§ 1522(b)(1)), primarily relies upon an ADS output (§ 1522(b)(2)), uses an ADS output (§ 1522(c)), has primarily used an ADS (§ 1522(d)) or primarily relied upon an ADS (§ 1524). Except as §§ 1526.5 and 1526.6 provide. Part 5.5.5 becomes operative on July 1, 2027 (§ 1526.7).

Law · Pending

Colorado HB 26-1139

State of Colorado

Binding

A carrier that uses an artificial intelligence system for the purpose of utilization review, or that contracts with or otherwise works through a person that does (§ 10-16-112.7(2)(a)); a pharmacy benefit manager or private utilization review organization that contracts with a carrier to provide utilization review services on behalf of the carrier and uses an artificial intelligence system to conduct the utilization review (§ 10-16-112.7(2)(b)); a behavioral health administrative services organization or managed care entity that uses an artificial intelligence system for the purpose of conducting utilization review of mental or behavioral health services (§ 10-16-112.7(2)(c)); a carrier offering a health benefit plan issued or renewed in the state on or after the effective date of the section (§ 10-16-112.7(6)(a)); and a payer of mental or behavioral health-care services provided under the Colorado Medical Assistance Act or the Children's Basic Health Plan Act (§ 25.5-1-209). The Act takes effect January 1, 2027 (Act § 4(1)), subject to the referendum-petition contingency in § 4(1).

Law · Pending

E2SSB 5395

State of Washington

Binding

An artificial intelligence system that determines or influences prior authorization for a Washington health plan, deployed by a carrier offering a health plan issued or renewed on or after January 1, 2024 (RCW 48.43.830(1)), by a health plan offered to public employees, retirees and their covered dependents under chapter 41.05 RCW (RCW 41.05.845(1)), or by a health care benefit manager as defined in RCW 48.200.020 carrying out prior authorization functions under direct or indirect contract with a carrier (RCW 48.43.830(6)). A provider or facility submitting requests is outside the act.

Law · In Force

Ind. Code ch. 27-1-52 (Downcoding of Health Benefits Claims)

State of Indiana

Binding

Indiana health insurers and health maintenance organizations — may not use automated systems or artificial intelligence as the sole basis to downcode a health claim based on medical necessity without a human reviewing the beneficiary medical record; healthcare providers may not use automated tools to submit claims without review by the provider or billing professional.

Law · In Force

N.Y. State Technology Law § 402-a

State of New York

Binding

A county, city, town, village, school district, board of cooperative educational services, county vocational education and extension board, district corporation, the State University of New York, the City University of New York or a community college (§ 401(3)) that utilizes an automated employment decision-making tool, defined in § 401(1) as software that uses algorithms, computational models, or artificial intelligence techniques, or a combination, to materially automate or replace human decision-making regarding employment. A state agency is not a covered entity; § 402 states its disclosure duty.

Law · In Force

Oklahoma Responsible Technology in Schools Act (S.B. 1734)

State of Oklahoma

Binding

An Oklahoma public school district's use of classroom or student-facing AI tools, and any AI vendor whose tools are used by or disclosed by an Oklahoma public school district to parents/guardians.

Law · In Force

US Federal4 laws

All Emergency Medicine AI Summit — Statement of Principles v2.0

United States (voluntary)

Guidance

Emergency departments deploying AI for triage, diagnosis, documentation, or operational decision support.

Guidance · In Force

CMS AI Playbook v4 — Auditable Data Lineage and Oversight

United States (federal)

Guidance

Use of AI in workflows that influence Medicare or Medicaid claims, prior authorization, or coverage decisions.

Guidance · In Force

False Claims Act

United States (federal)

Binding

Submission of false or fraudulent claims for payment to the U.S. government, including by AI-influenced clinical documentation and billing systems.

Law · In Force

FINRA 2026 Annual Regulatory Oversight Report — generative artificial intelligence examination themes

United States (Federal)

Guidance

FINRA member firms using generative AI or autonomous AI agents.

Guidance · In Force

Americas3 laws

Canada ADM Directive

Canada (Federal)

Guidance

An institution subject to the Policy on Service and Digital — a department within the meaning of s. 2 of the Financial Administration Act — using an automated decision system in production to make an administrative decision or a related assessment about a client (Directive, ss. 5.1 and 8.1). Systems used solely for research and experimentation and those in test environments are out of scope (s. 5.2). The instrument reaches systems developed or procured after 1 April 2020, and systems predating that date once significantly modified (s. 1.2).

Guidance · In Force

EDSTA

Province of Ontario, Canada

Binding

The organization is a public sector entity within EDSTA s. 1 (1): an institution within the meaning of s. 2 (1) of the Freedom of Information and Protection of Privacy Act other than the Assembly, an institution within the meaning of s. 2 (1) of the Municipal Freedom of Information and Protection of Privacy Act, a children aid society, or a school board. The cyber duties in O. Reg. 51/26 attach only to the narrower class prescribed by its s. 2: FIPPA educational institutions, Group A, B and C public hospitals, the University of Ottawa Heart Institute, children aid societies and school boards. The under-18 notice duties in O. Reg. 52/26 attach to every school board. The artificial-intelligence duties in ss. 5 and 6 attach only to entities and circumstances prescribed under s. 7, and none is prescribed.

Law · In Force

Law on Promotion of Artificial Intelligence Use (Law No. 31814) and Regulation (Supreme Decree No. 115-2025-PCM)

Republic of Peru

Binding

Developers and deployers of AI systems in Peru's public administration, state companies, and private sector (excluding personal use and national defense/security).

Law · Partially In Force

Global & Voluntary2 laws

ISO/IEC TR 24027

International (voluntary)

Guidance

Surfaced as voluntary standards context wherever an AI system informs or makes a decision, at any lifecycle phase from data collection onwards. ISO/IEC TR 24027 creates no duty of its own and contains no requirements; where an organization is bound to use it, the binding instrument is a contract, a tender condition or an internal policy, and the exposure for a bias failure belongs to the instrument that governs the decision.

Standard · In Force

UNESCO Recommendation on the Ethics of AI

United Nations (UNESCO)

Guidance

Surfaced wherever an AI system is designed, developed, deployed or used, in any sector and for any purpose: para. 2 (b) defines AI actors as any natural or legal person involved in any stage of the life cycle, para. 4 offers ethical guidance to all AI actors including the private sector, and the Recommendation carves out no domain. It creates no duty on anyone — Member States apply it voluntarily and AI actors receive guidance. Three rows carry the condition their paragraph turns on: decisions about people (para. 38, decision limb) and the provider role (paras. 51, last sentence, and 120, last sentence).

Guidance · In Force

EU & EEA1 law

EIOPA Opinion on AI governance and risk management (EIOPA-BoS-25-360, 6 August 2025)

European Union

Guidance

Insurance undertakings and insurance intermediaries, insofar as they may use AI systems within their respective areas of competence in the insurance value chain (§ 1.3), other than AI systems that are prohibited AI practices or high-risk under the AI Act (§ 2.6).

Guidance · In Force

Switzerland1 law

revFADP (SR 235.1)

Switzerland

Binding

Processing of personal data of individuals in Switzerland, regardless of where the controller is established.

Law · In Force

Asia Pacific1 law

Uzbekistan AI Law (LRU-1115)

Republic of Uzbekistan

Binding

Any AI system reaching a person in Uzbekistan -- human oversight is required for legally significant decisions relating to a person's rights and freedoms, a no-harm standard applies to AI-created resources and AI-based systems, and AI-specific administrative liability attaches to unlawful personal-data processing or dissemination.

Law · In Force

Sourced from the LegisGate™ intelligence library — our single source of truth (SSOT). Theme membership is computed deterministically from curated library text; run a LegisGate™ report to see how these obligations apply to your specific AI tool.

Talk to usWe're here to help