California has enacted a new law requiring healthcare providers using artificial intelligence to keep clinical decision-making human.
AB 1979, signed by Governor Gavin Newsom on September 30, 2026, requires a health facility, clinic, physician’s office, or office of a group practice to take reasonable steps to ensure that a licensed healthcare provider, acting within the provider’s scope of practice, retains the ability to exercise independent professional judgment when patient care is informed by a clinical decision support system.
The law also prohibits these healthcare entities from using or deploying a tool, system, or device that includes AI to independently perform a clinical function that the law requires to be performed by a person with a professional license.
AB 1979 does not define “independent professional judgment and leaves this open to interpretation. This is similar to the approach the California legislature took in, SB 574, the new California law which prohibits lawyers from delegating the practice of law to generative AI. That law too, leaves the term “delegate the practice of law” undefined.
However, the opening left by this undefined term can be filled by the growing body of requirements for meaningful human intervention under laws including the Colorado Privacy Act, the California Consumer Privacy Act automated decision-making technology regulations, and regulatory guidance under the European GDPR which can provide some guideposts.
Based on these approaches, in order to retain decision-making and independence, the human should have both the theoretical and practical ability to override the AI and make the decision. It is not enough for a licensed professional to appear somewhere in the process or to approve a decision the AI has, for all practical purposes, already made.
Especially important are the professional’s legal and actual ability to:
- Evaluate the relevant patient information;
- Assess whether the AI output is clinically appropriate;
- Accept, reject, modify, or disregard the output;
- Select an alternative course of action; and
- Exercise the expertise and judgment required by the professional’s license.
This requires more than the appropriate knowledge, training, and skill set. The professional must also have the time, information, authority, and resources necessary to critically evaluate the AI output. The professional should not face retaliation or other adverse consequences for disagreeing with the system.
The central question is not simply whether a human is “in the loop.” It is whether the licensed professional retains meaningful control over the clinical decision.
Bottom line:
As organizations design the human-in-the-loop component of their AI governance programs, they should focus not merely on where a human appears in the process, but making sure that the human is not a “rubber stamp” and has the actual authority, expertise, information, time, and resources necessary to exercise meaningful independent judgment over the AI.
AB 1979 takes effect January 1, 2027.
Recent Comments