AB-2575 Health care services: artificial intelligence

What happened

The September 4 enrolled AB 2575 would prohibit retaliation or discrimination based solely on a direct-care worker’s override of, or reliance on, clinical decision-support output when using professional judgment within scope. It preserves duties concerning care standards, licensing, independent judgment and lawful operating policies; a policy requiring deference to AI would not count as such a duty. The Labor Commissioner could investigate and enforce retaliation protections, without its determination deciding malpractice or professional-license proceedings. These changes were proposed, not enacted, on September 5.

The proposal also would prevent a defendant that developed, modified, selected or deployed a clinical decision-support system from arguing that a worker’s failure to override its output was a superseding cause severing liability for alleged harm. It does not foreclose other causation/foreseeability defenses, comparative-fault evidence or evidence of reasonable precautions and disclosures. The provision excludes the specified professional-negligence actions against licensed/certified direct-care practitioners. These limits distinguish a particular defense restriction from automatic vendor liability or immunity for clinical decisions.

Why it matters

Organizations would need both employment protections and carefully scoped AI risk allocation while retaining clinical accountability.

Sources

• AB-2575 Health care services: artificial intelligence, bill historyOfficial · California Legislature · 2026-08-31 · Source review 2026-09-10
• AB-2575 Health care services: artificial intelligence, current bill text (09/04/26 - Enrolled)Official · California Legislature · 2026-09-04 · Source review 2026-09-10