If you read that Arizona banned AI from denying medical claims, you read the headline, not the law. Arizona HB 2175 took effect July 1, 2026, and NBC, the Arizona Medical Association, and a stack of compliance trackers all describe it as a ban on AI in claim review. The enacted text does not contain the phrase “artificial intelligence” anywhere. This is a small case study in why reading the statute beats reading the coverage.
What the law actually requires
HB 2175 adds two sections to Arizona’s insurance code governing who may deny a medical-necessity claim or prior authorization, and what they may rely on. In plain terms, a qualified human has to make and own the adverse determination, using independent professional judgment. It does not matter whether an AI tool helped reach it; a person with the right authority has to make the call. The statute is technology-neutral. It neither bans nor mentions AI, and a payer can use AI heavily and still comply, as long as the human-judgment requirement is met on a denial.
What it does not do
There is no new penalty, no private right of action, and no unprofessional-conduct provision, that limb was in the House version and did not survive. Enforcement sits with the Arizona Department of Insurance and Financial Institutions under its general authority.
Why the myth is expensive in both directions
Believe the headline and under-comply: a payer told “we just can’t use AI” can miss the actual duty, which is about human ownership of the denial, not the tool. You can violate the real rule while using no AI at all. Believe the headline and over-comply: a payer could spend real money ripping AI out of a workflow to satisfy a prohibition that does not exist, while the actual requirement, a human in the loop on denials, goes unbuilt.
What to do now
1. Read the enacted text, not the tracker. On a fast-moving topic, the introduced bill and the enacted law are often different laws. 2. If you touch Arizona health claims: make sure a qualified human owns every medical-necessity denial, with independent judgment on the record. That is the duty. 3. Do not rip out AI to satisfy a ban that is not there. Keep the human in the consequential decision instead. 4. Apply the same discipline everywhere. A strong headline over a narrower enacted text is a pattern across state AI law.
An honest limitation
We are describing the enacted Arizona statute (HB 2175, Chapter 165), in force since July 1, 2026, read against its own text. Implementing guidance from the Department, if any, was not reviewed here, and the human-review duty has boundaries a payer should confirm against the statute and counsel before relying on a specific reading. The durable point stands: the law requires a human to own the denial; it does not ban AI.
This briefing is general information from Sentinel Assurance Group, not legal advice. Regulatory dates and requirements change — we maintain these briefings, but verify against primary sources and counsel before acting. Last reviewed August 17, 2026.
See how a Gap Assessment maps your exposure →Running AI in claims or utilization review?
Know where a human must own the call.
The free AI Risk Exposure call maps your health-AI workflows to the human-in-the-loop duties now spreading across the states.
Book the call →