“Clinician Should Have Overridden It” Defense Is Gone: New Liability Rules for Clinical AI Systems

Newsletter

The clinician didn’t catch the error, but that may no longer be a valid defense in California. California AB 2575 eliminates one of the most common shields artificial intelligence (AI) developers and deployers raise when a clinical decision support system (CDSS) causes patient harm: the argument that a licensed professional’s failure to override the AI’s output is a superseding cause that breaks the chain of liability. At the same time, the bill protects health care workers who exercise professional judgment to override, or rely on, AI outputs, barring employers from retaliating against them for either choice.

Notably, earlier versions of AB 2575 included transparency and disclosure requirements that would have required health care providers to make available information about how CDSS tools are used and trained, which were removed during the amendment process. The enrolled version of the bill focuses exclusively on liability allocation and worker anti-retaliation protections. However, even though your organization won’t be required to provide those answers upon patient request in the clinical setting, it should be prepared to do so should litigation arise.

This is the third installment in our series on California’s 2026 legislative sessions as it relates to the use of AI in health care. Each installment covers a single bill; this one addresses AB 2575, which pairs employment protections with the elimination of a common affirmative defense in tort litigation.

This installment analyzes the key provisions of AB 2575, outlines the compliance obligations it creates for affected stakeholders, and provides practical guidance for AI developers and deployers, as well as employers in the health care space.

In This Series
Part Three: New Liability and Worker Protection Framework for AI-Driven Clinical Decision Support Systems (AB 2575)

What it does: AB 2575 offers a two-fer in California’s evolving AI regulatory landscape: it reshapes the tort liability landscape for CDSS-related injuries and establishes employment protections for health care workers interacting with AI systems.

As we addressed in our preceding alert on SB 503, a CDSS is an AI system that produces a prediction, classification, recommendation, evaluation, or analysis that aids clinical decision-making related to timing of care, diagnosis, or treatment. The definition excludes booking, canceling and rescheduling appointments, appointment reminders, patient education and previsit materials and preparation, and payment processing, to the extent the independent performance of these activities by the system does not require a professional license.

As litigation against developers and deployers of AI tools increases, defenses feature a common refrain: even if the defendant was negligent, they can’t be held liable for the ultimate harm because an independent, unforeseeable intervening event broke the chain of causation between their conduct and the plaintiff’s injury. This is the superseding cause defense, and if AB 2575 becomes law, deployers and developers would be barred from asserting that a health care professional’s failure to override the system’s output was a superseding cause that severs the defendant’s liability.

The bill does not, however, eliminate all defenses. Defendants may still present: (1) any other affirmative defense, including evidence relevant to causation or foreseeability; (2) evidence relevant to the comparative fault of any other person or entity; and (3) evidence that the defendant took reasonable precautions to prevent harm, including providing clear and conspicuous disclosures and disclaimers regarding the intended use, scope, known risks, and limitations of the system to the health facility, clinic, physician’s office, or group practice. Nor does the bill’s civil liability provision apply to professional negligence actions against a person providing direct patient care who hold specific licenses or certifications (e.g., physicians, chiropractors, dentists, pharmacists, PAs, DOs).

AB 2575 also creates new employment protections enforceable under the Labor Code. The bill prohibits employers from retaliating or discriminating against a worker providing direct patient care using their professional judgment to make an assessment or decision within their appropriate scope of practice based solely on the worker’s override of, or reliance on, the output of a CDSS. Notably, a policy, procedure, or protocol that requires a worker to accept, defer to, or refrain from overriding the output of a CDSS is not a protected employer duty under the bill. This means employers cannot use internal policies to compel blind adherence to AI recommendations and then claim the worker violated policy.

The anti-retaliation provision and the limitation on superseding-cause defenses will also have significant implications for health care malpractice litigation in cases involving AI-assisted care. For example, a pharmacy that develops clinical decision support tools, or selects and deploys them, is both likely to be obligated to protect its workers when they reject an output and face a liability limitation in malpractice or negligence suits.

Key compliance obligations:

  • Employers are prohibited from retaliating or discriminating against workers providing direct patient care who use their professional judgment to make an assessment or decision within their appropriate scope of practice based solely on the worker’s override of, or reliance on, the output of a CDSS.
  • Anti-retaliation protection does not affect a worker’s duty to meet the applicable standard of care, act within their scope of practice, exercise independent professional judgment, comply with standard operating procedures or other written policies established by the employer/regulator, or comply with certain regulatory requirements.
  • Workers subject to retaliation or discrimination may file a complaint with the Labor Commissioner, who may investigate, order temporary relief, issue citations, or file civil action.

Effective date: January 1, 2027.

Practical takeaways for CDSS developers and deployers: Although the enrolled bill has eliminated the transparency and documentation provisions, if the bill becomes law, organizations that develop, modify, select, or deploy clinical decision support systems will still face documentation imperatives based on potential litigation exposure.

  • Audit Existing Disclosures. Review all current disclosures, disclaimers, and documentation provided to health care clients. Ensure they clearly articulate the intended use, scope, known risks, and limitations of each CDSS product. Vague or boilerplate disclaimers are unlikely to support a “reasonable precautions” defense.
  • Strengthen Risk Documentation. Develop and maintain comprehensive documentation of system design, training data, validation processes, known limitations, and risk assessments. Where that information sits with the vendor, obtain it through diligence and secure it contractually, including an affirmative obligation for the vendor to provide material updates. This documentation will be critical to establishing the “reasonable precautions” defense in potential litigation.
  • Revisit Contractual Risk Allocation. Review and update contractual agreements to address the new liability framework, including indemnification provisions, insurance requirements, and representations about system capabilities and limitations.
  • Ongoing Monitoring and Updates. Because the “reasonable precautions” defense may encompass ongoing conduct, developers and deployers should implement post-deployment monitoring, incident tracking, and timely communication of newly discovered risks or limitations.
  • Prepare for Litigation. Reassess litigation risk allocation with counsel and insurers, recognizing that a defendant that developed, modified, selected, or deployed a clinical decision support system cannot rely on a clinician’s failure to override an output as a superseding cause. Revisit indemnity, defense, and limitation of liability terms in vendor agreements accordingly.

Practical takeaways for health care employers using CDSS: Hospitals, clinics, pharmacies, physician groups, and other health care employers should be ready to adapt employment practices to the new anti-retaliation framework:

  • Conduct a Policy Audit. Review employment policies, procedures, and protocols to ensure that none require workers to accept, defer to, or refrain from overriding the output of a CDSS. Any such policy is explicitly prohibited and could form the basis of a retaliation claim.
  • Refresh AI Governance Frameworks. Update internal governance structures for AI/CDSS procurement, deployment, and monitoring. Include clinician input in CDSS selection and evaluation processes.
  • Implement Anti-Retaliation Training. Train managers, supervisors, and leadership on anti-retaliation requirements to clarify that adverse employment actions cannot be based solely on a worker’s decision to override AI recommendations. Incorporate scenario-based training to address gray areas.
  • Documentation Decision-Making. Employers should document the reasons for any adverse employment action taken against a worker who has overridden CDSS output. The “based solely on” standard means employers may still discipline workers for failing to meet the standard of care, acting outside their scope of practice, or violating legitimate operational policies—but the employer bears the burden of demonstrating that the action was not solely based on the override.
  • Labor Commissioner Complaint Preparedness. Establish internal processes to respond to Labor Commissioner investigations, including document retention policies and designated points of contact.      

In sum, AB 2575 signals a legislative intent to ensure that AI integration into clinical care does not erode either patient safety or clinician autonomy. Part Four of this series takes up SB 903, which applies these same themes of human authority and informed consent to a single practice area: the delivery of psychotherapy services.

For any questions on procurement or deployment of CDSS or related employee AI policies in the health care setting contact your Quarles attorney or:

Follow Quarles

Subscribe Media Contact
Back to Main Content

We use cookies to provide you with the best user experience on our website and to analyze statistics related to our website. To understand more about how we use cookies, or for instructions to change your preference and browser settings, please see our Privacy Notice. Please note that if you choose to reject cookies, doing so may impair some of our website's functionality.