Right to Override Act
- Sponsor
- Sen. Markey, Edward J. [D-MA]
- Committees
- Health, Education, Labor, and Pensions Committee (primary)
- Last action
- Oct 9, 2025
Bottom line
The 'Right to Override Act' aims to safeguard the independent clinical judgment of healthcare professionals in the face of increasing AI integration, providing explicit protections against retaliation and mandating structured oversight for AI use in healthcare settings.
What it actually does
This bill establishes federal protections for healthcare professionals who override Artificial Intelligence Clinical Decision Support System (AI/CDSS) outputs. It requires healthcare entities to implement policies that ensure professional judgment is not supplanted by AI, allows timely overrides, provides training, and mandates the creation of AI/CDSS committees. The bill also prohibits adverse employment actions or retaliation against professionals for overriding AI/CDSS outputs and establishes enforcement mechanisms through the Departments of Health and Human Services and Labor, including civil monetary penalties and a private right of action.
Proponents argue
Proponents argue that the bill is essential for patient safety, ensuring that human oversight remains paramount in healthcare decisions, especially as AI technologies become more prevalent. They contend it protects the professional autonomy of doctors, nurses, and other healthcare providers, preventing potential algorithmic biases or errors from leading to suboptimal care. Furthermore, supporters believe it empowers professionals to provide feedback on AI systems, leading to better, more reliable technology over time.
Opponents contend
Opponents might argue that while the intent is good, the bill could create significant administrative burdens and compliance costs for healthcare entities, potentially slowing the adoption of beneficial AI technologies. Concerns could also be raised about the potential for increased liability for individual providers if overrides lead to adverse outcomes, or if the bill's provisions create ambiguity regarding accountability. Some may also suggest that overly prescriptive regulations could stifle innovation in AI development for healthcare.
The bill is relatively concise and clearly structured, making it accessible for an informed reader to understand its core provisions and implications within a reasonable timeframe.
Sec. 101(a)(1)
Mandatory Policies for AI/CDSS Usage
This provision requires covered healthcare entities that use AI/CDSS to adopt and adhere to a policy ensuring that AI outputs do not replace the independent judgment of healthcare professionals. The policy must allow professionals to override AI/CDSS outputs in a timely manner if deemed appropriate for the patient or necessary to comply with applicable law, including civil rights law.
Supporters argue
Supporters argue this provision is crucial for patient safety, ensuring that human clinical judgment, which can account for nuanced patient factors and ethical considerations, always takes precedence over algorithmic recommendations. It prevents the 'automation bias' where professionals might blindly follow AI outputs.
Critics contend
Critics might contend that this provision, while well-intentioned, could create a disincentive for healthcare entities to fully integrate AI, or could lead to inconsistent application of AI benefits if overrides become too frequent without clear justification, potentially undermining the efficiency and data-driven insights AI offers.
Tradeoffs
This provision balances the efficiency and data-driven insights offered by AI with the imperative to maintain human accountability and ethical judgment in healthcare. The tension lies in defining the precise circumstances where an override is 'appropriate' without creating undue liability or administrative friction.
Sec. 101(a)(3)
Mandatory AI/CDSS Training for Professionals
Covered entities are required to provide comprehensive training to healthcare professionals on how to use AI/CDSS, when an override is appropriate, how to perform an override, the AI's development processes and data inputs, and any potential limitations or biases of the AI system.
Sec. 101(a)(4)
Establishment of AI/CDSS Committees
This provision mandates that covered entities establish an AI/CDSS committee. These committees must convene shortly after AI/CDSS adoption or enactment of the Act, be composed of at least as many non-managers as managers, include labor organization representatives, and meet quarterly to review policies, provide consultation, and suggest improvements.
Sec. 201
Prohibition on Adverse Employment Actions
This section explicitly prohibits covered entities from taking any adverse employment action against a healthcare professional solely because that professional overrides an AI/CDSS output in a manner consistent with the Act's requirements. This provides a direct protection against retaliation for exercising independent judgment.
Sec. 203(b)
Private Right of Action and Remedies
This section grants individuals adversely affected by a violation of Title II (adverse employment actions, whistleblower protections) the right to file a civil action in federal court. Successful plaintiffs can receive various forms of relief, including actual damages (up to treble damages), statutory damages ($5,000-$20,000 for adverse actions, $10,000-$100,000 for whistleblower violations), injunctive relief, equitable relief, and attorney fees. It also includes a waiver of sovereign immunity for states receiving federal financial assistance.
The definition of 'adverse employment action' is exceptionally broad, encompassing not only termination or demotion but also disciplinary actions, burdensome work schedules, denial of promotions or benefits (including telework/training), revocation of admitting privileges, duty reassignments, adverse evaluations, and any other modification or inaction that puts a professional in a 'materially adverse position'.
Sec. 3(1)
Why it matters:The breadth of this definition is standard legislative practice to ensure comprehensive protection against various forms of retaliation, leaving little room for loopholes. It aims to cover both direct and indirect forms of adverse actions.
Case for: Supporters would argue that this expansive definition is necessary to genuinely protect healthcare professionals from subtle or indirect forms of retaliation that might not be covered by a narrower definition, ensuring the spirit of the Act is upheld.
Case against: Critics might argue that such a broad definition could lead to ambiguity and make it challenging for covered entities to manage personnel, potentially opening the door to disputes over minor workplace changes or perceived slights, increasing administrative and legal burdens.
Estimated impact: Significantly expands the scope of protected employment actions, increasing legal risk for covered entities if they are found to have retaliated against healthcare professionals for overriding AI/CDSS outputs.
The definition of 'engaged in the performance of work for remuneration' explicitly includes individuals having 'admitting privileges' for a covered entity, regardless of whether they are directly employed by that entity.
Sec. 3(6)
Why it matters:This is likely a deliberate legislative choice to ensure that independent practitioners who are not direct employees but have significant working relationships with covered entities (e.g., hospital-affiliated physicians) are also covered by the Act's protections, preventing a loophole.
Case for: Proponents would argue that including those with admitting privileges is crucial for comprehensive protection, as these professionals also rely on AI/CDSS within the entity's system and face similar pressures regarding independent judgment.
Case against: Opponents might argue that extending protections to non-employees could complicate the legal relationship between entities and independent contractors, potentially creating new compliance challenges and blurring lines of responsibility.
Estimated impact: Expands the reach of the Act's protections to a broader category of healthcare professionals who are not direct employees but are affiliated with covered entities, increasing the number of individuals and relationships subject to the Act.
The bill generally prohibits sharing 'override data' on specific health care professionals or groups where identity can be inferred. However, it includes critical exceptions: this prohibition does not apply when informing a patient or their authorized representative about a care decision, or in cases of civil, criminal, or administrative action involving medical malpractice, negligence, or violation of any law.
Sec. 101(a)(1)(D) and Sec. 101(b)
Why it matters:This structure is a common legislative technique to state a general rule and then carve out specific, necessary exceptions. The exceptions are crucial for maintaining transparency with patients and ensuring accountability in legal contexts, balancing privacy with other critical needs.
Case for: Proponents would highlight that these exceptions are vital for patient transparency and legal accountability. They ensure that while professionals are protected from unwarranted scrutiny, legitimate needs for data sharing in patient care and legal proceedings are preserved.
Case against: Critics might argue that even with exceptions, the general prohibition on sharing individual override data could hinder internal quality improvement efforts or the ability to identify patterns of problematic overrides that might indicate a need for further training or system adjustments, outside of formal legal action.
Estimated impact: Creates a framework where individual override data is generally protected but can be accessed for specific, critical purposes related to patient care and legal accountability, influencing data governance and privacy practices within healthcare entities.
The bill includes a non-preemption clause, stating that nothing in the Act shall preempt a State law or collective bargaining agreement.
Sec. 304
Why it matters:Non-preemption clauses are common in federal legislation, particularly in areas where states have existing regulatory frameworks or where labor rights are involved. Its placement at the end is standard for general provisions.
Case for: Proponents would argue that this clause is crucial for allowing states to enact stronger protections or for collective bargaining agreements to provide additional benefits, ensuring the Act serves as a floor, not a ceiling, for rights and protections.
Case against: Critics might argue that a non-preemption clause could lead to a patchwork of differing state laws and collective bargaining agreements, potentially creating compliance complexities for healthcare entities operating across state lines or with multiple labor contracts.
Estimated impact: Preserves the ability of states and labor organizations to establish additional or more stringent protections related to AI/CDSS overrides and whistleblower rights, potentially leading to varied regulatory landscapes across jurisdictions.