Doctors Not AI Act of 2026
- Sponsor
- Rep. Landsman, Greg [D-OH-1]
- Committees
- Education and Workforce Committee (primary)
- Last action
- Sep 1, 2026
Bottom line
The bill aims to safeguard human clinical judgment in health care coverage decisions, ensure transparency in AI use, and extend mental health parity protections to AI-driven utilization review processes.
What it actually does
This bill prohibits artificial intelligence (AI) systems from making or dictating adverse health care benefit determinations that involve clinical judgment. It mandates that such decisions must be made by licensed health care professionals who exercise independent clinical judgment, without deferring to AI outputs. The bill also requires transparency regarding the use of AI in claims review and explicitly subjects AI systems used in utilization review to mental health parity laws.
Proponents argue
Supporters argue that the bill protects patients from potentially biased, inaccurate, or opaque AI-driven denials of care, ensuring that complex medical decisions are made by qualified human professionals. They contend it promotes transparency and accountability in a rapidly evolving technological landscape, preventing 'black box' denials and upholding the critical role of the doctor-patient relationship.
Opponents contend
Opponents might argue that the bill could stifle innovation in health technology, increase administrative costs for health plans and issuers, and potentially slow down claims processing. They may contend that AI, when properly designed and overseen, can improve efficiency, consistency, and accuracy in claims review, potentially reducing human error and administrative burdens.
The bill is concise and clearly written, allowing an informed reader to quickly understand its core intent and operative provisions.
Sections 2(a), 3(a), 4(a)
Definition of Artificial Intelligence System
This provision defines an 'artificial intelligence system' broadly. It includes any engineered or machine-based system, algorithm, predictive model, machine learning system, or automated decision-making software that processes data to generate predictions, classifications, recommendations, or decisions, and is used to materially influence, automate, or issue determinations regarding health care coverage.
Supporters argue
Proponents argue that a broad definition is necessary to ensure comprehensive oversight of all forms of automated decision-making that could impact patient care, preventing loopholes as technology evolves.
Critics contend
Opponents might argue that an overly broad definition could inadvertently capture simple automation tools, leading to unnecessary regulatory burdens and stifling innovation in beneficial health tech applications.
Tradeoffs
The tension lies between ensuring comprehensive regulatory coverage of AI technologies and avoiding over-regulation that could impede the development and adoption of beneficial automated tools in healthcare.
Sections 2(b), 3(b), 4(b)
Prohibition on AI-Dictated Adverse Benefit Determinations
This core provision stipulates that adverse benefit determinations involving clinical judgment cannot be issued or dictated by an artificial intelligence system. It mandates that these decisions must be made by a licensed health care professional acting within their scope of practice and with relevant training and experience.
Sections 2(b)(3), 3(b)(3)(C), 4(b)(3)
Requirement for Independent Clinical Judgment
This section requires that the licensed health care professional making an adverse benefit determination involving clinical judgment must reflect their independent clinical judgment. They must conduct an independent evaluation of the enrollee's medical circumstances and are explicitly forbidden from treating any output of an AI system as presumptively valid or deferring to it.
Sections 2(b)(4), 3(b)(3)(D), 4(b)(4)
Transparency in AI Use for Notices
If an artificial intelligence system was used in connection with an adverse benefit determination, the notice provided to the enrollee must include a statement that such system was used, a description of its role in the review process, and the name, professional license, and credentials of the licensed health care professional who made the determination.
Sections 2(b)(5), 3(b)(3)(E), 4(b)(5)
Documentation Requirements for AI Use
The bill requires that documentation be maintained as part of the administrative record describing the artificial intelligence system used, its role in the review process, any outputs, scores, recommendations, or determinations generated by the system, and documentation demonstrating compliance with the independent clinical judgment requirement. These materials must be made available to the enrollee upon request.
Sections 2(c), 3(c), 4(c)
AI Systems as Treatment Limitations for Mental Health Parity
This provision amends existing mental health parity laws to explicitly state that the use of an artificial intelligence system in connection with utilization review constitutes a 'treatment limitation.' It also requires plans and issuers to provide sufficient information regarding the function, operation, and effects of such AI systems to enable regulators to evaluate their compliance with mental health and substance use disorder parity requirements.
The definition of 'artificial intelligence system' is broad, encompassing 'any algorithm, predictive model, machine learning system, or automated decision-making software' that processes data to generate predictions, classifications, recommendations, or decisions and is used to 'materially influence, automate, or issue determinations' regarding health care coverage.
Sections 2(a), 3(a), 4(a) (Definition of Artificial Intelligence System)
Why it matters:This is standard legislative practice to establish the scope of a bill's applicability. However, its broadness is critical to the bill's reach and could be seen as a strategic choice to future-proof the legislation against evolving AI technologies.
Case for: A broad definition ensures that the regulations apply comprehensively to various forms of automated decision-making in healthcare, preventing circumvention as technology advances. It aims to cover the spirit of automated influence, not just specific current AI models.
Case against: Critics might argue that such a broad definition could inadvertently capture simple, non-AI automation tools, leading to over-regulation and compliance burdens for systems that pose minimal risk to clinical judgment.
Estimated impact: This definition dictates the entire scope of the bill's regulatory impact on health plans and health tech developers, potentially affecting a wide array of existing and future automated systems.
The bill explicitly states that 'The use of an artificial intelligence system in connection with utilization review shall constitute a treatment limitation for purposes of this section' (referring to mental health parity sections).
Sections 2(c)(2), 3(h), 4(f) (Parity Treatment of Artificial Intelligence Systems)
Why it matters:This is a technical amendment to existing law, a common legislative drafting technique. However, its specific inclusion highlights a deliberate effort to address a potential loophole or emerging issue in mental health parity enforcement.
Case for: Proponents argue this ensures that AI algorithms cannot be used to create discriminatory barriers to mental health or substance use disorder treatment, aligning AI-driven utilization review with existing parity protections.
Case against: Opponents might argue that this adds significant complexity to mental health parity compliance, requiring health plans to conduct potentially difficult and resource-intensive analyses of algorithmic bias in utilization review.
Estimated impact: This provision could significantly impact how AI is developed and deployed for utilization review, particularly concerning mental health and substance use disorder benefits, potentially leading to increased scrutiny and adjustments in algorithmic design.