Health care services: artificial intelligence.

The AI clinical decision support provisions were relocated from Section 1339.77 of the Health and Safety Code to a new Chapter 25.5 (Section 22758.5) in Division 8 of the Business and Professions Code, changing the regulatory framework from health facility regulation to professional licensing.

The requirement that no clinical decision be based solely on AI output was removed. Replaced with a requirement that licensed providers retain the ability to exercise independent professional judgment when care is informed by clinical decision support system output.

Enforcement mechanisms for licensed facilities (Article 4 of Chapter 2, Sections 1280 and 1280.3) and licensed clinics (Article 4 of Chapter 1, Section 1229) were removed, along with the provision making violations constitute unfair competition under Section 17200 of the Business and Professions Code.

Violations by physicians are now subject to the jurisdiction of the Medical Board of California or the Osteopathic Medical Board of California. Licensing boards may pursue injunctions where a violation constitutes unlicensed practice.

The definition of 'clinical decision support system' was changed from 'an automated decision system or generative artificial intelligence system whose outputs are that produces a prediction...' (containing a drafting error) to 'an artificial intelligence system that produces a prediction, classification, recommendation, evaluation, or analysis that aids clinical decisionmaking related to timing of care, diagnosis, or treatment.' New exclusions were added for appointment management, patient education, and payment processing systems.

The definition of 'clinical decision' (assessment of patient condition and education of a patient or family concerning health care problems, including postdischarge care) was removed entirely.

The requirement for health facilities and clinics changed from 'shall ensure that no clinical decision is based solely on the output of a clinical decision support system' to 'shall take reasonable steps to ensure' the same. This lowers the standard from an absolute guarantee to a reasonableness standard, giving providers more flexibility in compliance.

The independent professional judgment requirement was broadened. Previously it applied when a professional was 'reviewing and approving a clinical decision that is based on' AI output. Now it applies 'in their care of a patient whenever that care is informed by' AI output, covering a wider range of clinical interactions.

A new prohibition was added: health facilities and clinics may not use or deploy AI to independently perform any clinical function that is required by law to be performed by a person with a professional license. Previously the bill only prohibited AI from directing, guiding, supervising, or instructing unlicensed personnel in such functions.

A new exemption was added: the AI prohibition does not apply to a trainee's use of a tool, system, or device as part of a supervised course of study or training program while working toward licensure.

A new enforcement provision was added making violations by physicians subject to the jurisdiction of the Medical Board of California or the Osteopathic Medical Board of California, as appropriate. Specific code section references (Sections 1280 and 1280.3 for health facilities, Section 1229 for clinics) were also added.

The definition of 'clinical decision support system' was narrowed to specify that the system must produce 'a prediction, classification, recommendation, evaluation, or analysis' used to inform clinical decisionmaking, rather than the broader prior language of just 'outputs are used to inform clinical decisionmaking.'

The definition of 'health care chatbot' was restructured from a single-sentence definition into a multi-part numbered definition with three required elements: (1) adaptive human-like responses, (2) marketed as facilitating or supporting health services, and (3) uses health care chatbot information to facilitate or support health services. The scope of covered services was also broadened from 'mental or physical health' to simply 'health services.'

The definition of 'health care chatbot' was restructured from a single paragraph into a numbered list and broadened: 'mental or physical health services' was changed to simply 'health services,' which could capture a wider range of health-related chatbots. The 'health care chatbot information' definition was also expanded to include information a consumer provides 'either directly or by allowing access to that information.'

The definition of 'clinical decision' in Section 1339.77(a)(3) was narrowed by removing the phrase about 'communication between licensed health care professionals relating to the handoff of responsibility for a patient, and other documentation and communication that requires the application of a licensed health care professional's professional expertise.'

The exemption in Section 1339.77(d) was broadened from 'artificial intelligence' to 'automated decision systems,' which is a defined term in the bill covering computational processes derived from machine learning, statistical modeling, data analytics, or AI. This means more types of automated documentation and communication tools are now explicitly exempt from the clinical decision support requirements.

Section 1339.77(a)(2) was clarified to remove the awkward phrase 'exercises professional' and now reads that a licensed health care professional, acting within their scope of practice, retains the ability to exercise independent professional judgment when reviewing AI-based clinical decisions.

Throughout the bill, 'healthcare' was corrected to 'health care' (two words) for consistency with California legislative style. Page numbering was corrected from 94 to 93.

Replaced the broad prohibition against using AI to replace professional judgment with a narrower requirement that no clinical decision be based solely on the output of a clinical decision support system, and that a licensed professional retains the ability to exercise independent professional judgment when reviewing such output.

Added 'acting within their scope of practice' and 'retains the ability to' language to the requirement that licensed professionals exercise independent judgment, clarifying that the standard applies within each professional's licensed scope.

Narrowed the prohibition on AI directing unlicensed personnel from any function that requires a professional license to only clinical functions that are required by law to be performed by a person with a professional license, excluding non-clinical tasks.

Removed the clarification that 'managing the individual's information' includes the ability to query medical history, summarize doctor's notes, or organize lab results, which had been added in the Assembly version to expand CMIA coverage of consumer health apps.

Fixed cross-reference errors: corrected 'subdivision (c), (e)' to 'subdivision (e)' in the pharmaceutical company definition, and corrected 'subdivisions (d) and (e) (e) and (f)' to 'subdivisions (e) and (f)' in the marketing exception for chronic conditions.

Changed 'healthcare chatbot' to 'health care chatbot' throughout for consistency with the bill's existing terminology convention.

A new Section 56.05 of the Civil Code is added, containing definitions for terms including 'healthcare chatbot,' 'healthcare chatbot information,' 'generative artificial intelligence,' 'marketing,' 'medical information,' 'sensitive services,' and 'immigration enforcement.' This broadens the definitional framework of the CMIA.

A new subdivision (f) is added to Section 56.06, deeming any business that offers a healthcare chatbot to a consumer for managing information or diagnosis, treatment, or management of a medical condition to be a provider of health care subject to CMIA confidentiality and penalty requirements.

Section 1339.77 now requires that no clinical decision be based solely on the output of a clinical decision support system, and that a licensed health care professional exercises independent professional judgment when reviewing and approving a clinical decision based on such a system's output.

Two new definitions are added to Section 1339.77: 'automated decision system' (a computational process using machine learning, statistical modeling, data analytics, or AI that issues simplified output used to assist or replace human discretionary decisionmaking) and 'clinical decision support system' (an automated decision system or generative AI system whose outputs inform clinical decisionmaking regarding provision, timing, or course of patient care).

The bill's section numbering was reorganized from 3 sections to 4 sections to accommodate the new Section 56.05 definitions provision. The Health and Safety Code section number was corrected from 1339.76 to 1339.77.

Section 56.05 of the Civil Code (a full definitions section with over 20 defined terms including 'artificial intelligence health model,' 'medical information,' 'marketing,' 'sensitive services,' 'immigration enforcement,' etc.) is entirely removed from the bill.

Subdivision (f) of Section 56.06, which would have deemed any business offering an artificial intelligence health model to be a provider of health care subject to CMIA requirements, is removed. This means AI health model businesses are no longer brought under the Confidentiality of Medical Information Act by this bill.

The core AI prohibition in health facilities is narrowed. The old language prohibited AI for 'any activity requiring the use of professional judgment by a licensed health care professional.' The new language prohibits AI for 'any activity requiring to replace the use of professional judgment by a licensed health care professional, professional in carrying out their duties.' This allows AI to assist with tasks requiring professional judgment as long as it does not fully replace the professional's judgment.

The AI use restriction provisions are renumbered from Section 1339.76 to Section 1339.77 of the Health and Safety Code. Section 1339.76 is added as an empty placeholder section.

Added new subparagraph (2) to Section 56.06(b) clarifying that 'manage the individual's information' includes the ability to query medical history, summarize doctor's notes, or organize lab results. The bill states this is intended to clarify existing law rather than expand it.

Removed duplicate subdivision letters in Section 56.05 definitions (the old version had repeated letters like (a), (a), (b), (b), etc.). The new version uses clean sequential lettering from (a) through (v).

Corrected grammatical error in the bill title from 'Section Sections 56.05 and 56.06' to 'Sections 56.05 and Section 56.06.'

Removed 'Confidentiality of Medical Information Act: definitions' from the legislative counsel's digest heading, leaving only 'Health care services: artificial intelligence' as the primary topic.

Fiscal committee vote changed from 'no' to 'yes' and state-mandated local program designation changed from 'no' to 'yes,' indicating the bill now carries a fiscal note and imposes costs on local agencies.

New definition of 'artificial intelligence health model' added to Civil Code Section 56.05: an AI model (as defined in Business and Professions Code Section 22757.11) that offers individuals the option to securely connect or integrate their medical record for understanding and managing health and wellness.

New subdivision (f) added to Civil Code Section 56.06 deeming any business that offers an artificial intelligence health model to be a 'provider of health care' under CMIA, subjecting it to the same confidentiality standards and penalties for improper use or disclosure of medical information as traditional health care providers.

Explicit carve-out in new Section 1339.76: AI use for documentation and communication that does not involve the application of professional judgment (such as automated messages informing patients of health record updates) is still permitted.

Entirely new Health and Safety Code Section 1339.76 prohibits health facilities, clinics, physician's offices, and group practices from using or deploying AI for any activity requiring professional judgment by a licensed health care professional, including patient condition assessment, educating patients about health care problems (including postdischarge care), and communication between professionals regarding patient handoff.

New Section 1339.76 also prohibits using AI to direct, guide, supervise, or instruct unlicensed personnel in performing any function that requires a professional license.

Multiple enforcement mechanisms added for violations of new Section 1339.76: existing facility and clinic enforcement articles apply; violations constitute 'unfair competition' under Business and Professions Code Section 17200; and professional licensing boards may pursue injunctions or restraining orders where a violation constitutes unlicensed practice of a health care profession.

Fiscal committee review changed from 'no' to 'yes' and state-mandated local program changed from 'no' to 'yes,' reflecting that the bill now expands the scope of a crime and imposes costs on local agencies.

New Section 4 added providing that no state reimbursement is required under Article XIII B of the California Constitution because the only costs incurred by local agencies result from the creation or change of a crime.