AI Regulation Tracker / Legislation
Georgia Bars AI-Only Health Insurance Denials Under Signed SB 444
Georgia has signed Senate Bill 444, creating O.C.G.A. 33-46-7.1. It bars an AI system from issuing a health-coverage adverse determination until a clinical peer takes part in the review, and it forbids AI from overriding that peer's judgment. The bill is signed, but it does not take effect until January 1, 2027.
Georgia has signed Senate Bill 444, which creates a new section of the state insurance code, O.C.G.A. 33-46-7.1, dealing directly with the use of artificial intelligence in health-coverage decisions. The bill passed in the 2026 session and has been signed. The one thing to get straight before anything else is timing. This is law, but it is not operative yet. Section 2 of the bill reads, "This Act shall become effective on January 1, 2027." So the rule is settled, and the runway to comply runs to the end of this year.
What the law actually requires
The heart of SB 444 is a limit on what an AI system is allowed to do on its own in a coverage denial. The operative language is worth quoting in full, because the wording is what governs. The statute provides that an AI system "shall not issue an adverse determination to a patient until a natural person qualifying as a private review agent or a utilization review entity conducts a utilization review in which a clinical peer participates. In no event shall artificial intelligence systems ... supersede the judgment of such clinical peer."
Read that carefully, because it is precise. Two things have to happen before a denial can go out. First, a real person, one who qualifies as a private review agent or a utilization review entity, has to conduct a utilization review. Second, a clinical peer has to take part in that review. Only after both of those conditions are met can an adverse determination reach the patient. And even then, the AI cannot override the clinical peer. The peer's judgment controls.
An adverse determination is the industry term for a coverage decision that goes against the patient, a denial or a reduction of requested care. A clinical peer is a person with the clinical background to evaluate the specific care at issue. So the law is targeting the exact scenario that has drawn the most public concern: a model spitting out a denial with no qualified human actually weighing the case.
What the law does not do
It is just as important to be clear about the boundaries of this, because it is easy to overstate.
SB 444 does not ban AI from claims and utilization work. Insurers can still use AI to sort, flag, summarize, and assist the people doing reviews. What the law prohibits is letting AI be the one that issues the denial without a clinical peer in the loop, and letting AI override the peer once that peer has weighed in. This is a ceiling on AI's authority, not a prohibition on AI-assisted work.
The law also does not add a disclosure mandate. There is no new requirement here to tell patients that AI was involved, and no separate notice or labeling obligation created by this section. Some other states have gone that route. Georgia, in this bill, focused on the human-review requirement rather than on disclosure.
And, again, it is not yet in force. Nothing in a Georgia insurer's or reviewer's operations is legally required to change because of this section until January 1, 2027. Between now and then, the obligation is to be ready.
How this fits the national picture
Georgia is not acting alone here, and that context matters for anyone operating across state lines. SB 444 is part of a broader wave of state laws putting a human back in the loop on AI-driven claim denials. Arizona moved with HB 2175, Iowa with HF 2635, Indiana with HB 1271, and Washington with SB 5395. The details differ state to state, but the throughline is the same idea: a machine should not be the final word on denying someone's care without a qualified person accountable for the call.
For a national insurer or a utilization-review vendor, that pattern is the real signal. This is no longer a one-off. It is a set of parallel requirements landing in different states on different timelines, and the smart move is to design a compliant workflow once and apply it broadly rather than patching state by state at the last minute.
What this means for insurers and finance teams
If you run coverage operations touching Georgia, the work is concrete. You need to be able to show that no adverse determination goes out without a qualifying person conducting the utilization review and a clinical peer participating in it, and that your AI tools never override that peer. That is a process and documentation problem as much as a technology one. Who is the clinical peer on a given denial, what did they review, and can you prove the AI did not make the final call on its own.
For finance and compliance leaders sitting above the operation, this is a governance line item with a date on it. The effective date is January 1, 2027, which means the budget cycle, the vendor contracts, and the internal controls that carry you into 2027 are where this has to be built in. Waiting until the fourth quarter of next year to discover that your AI vendor's denial workflow does not have a clinical peer wired into it is the kind of gap that turns into regulatory exposure fast.
The practical thread is accountability. The law is asking a simple question that any auditor would recognize: for every denial, can you point to the qualified human who owned the decision. Build the answer into the process now, while the deadline is still a year out.
What to do now
Confirm whether your Georgia coverage operations rely on AI at the point of adverse determination, and map exactly where a human clinical peer sits in that flow today. Get the effective date on the calendar as January 1, 2027, and work back from it. Press your utilization-review vendors on how their AI tools handle denials, and get written confirmation that a clinical peer participates and that AI cannot supersede that peer. Build the documentation trail now so that, for any denial, you can name the qualifying reviewer and the participating clinical peer. And if you operate in more than one state, line SB 444 up against Arizona, Iowa, Indiana, and Washington and design one human-in-the-loop standard that clears all of them, rather than five separate fixes. As always, confirm how the statute applies to your specific operation with qualified counsel.
Questions professionals are asking
Is SB 444 in effect now?
No. The bill is signed, but Section 2 sets its effective date at January 1, 2027. It is settled law with a firm deadline, but it does not impose any legal obligation until that date.
Does SB 444 ban insurers from using AI?
No. It does not ban AI-assisted claims or utilization-review work. It bars an AI system from issuing an adverse determination to a patient until a qualifying person conducts a utilization review in which a clinical peer participates, and it forbids AI from overriding that clinical peer's judgment. AI can still assist; it just cannot be the accountable decision-maker on a denial.
Does the law require insurers to disclose when AI is used?
No. This section does not create a disclosure or notice mandate. It focuses on the human-review requirement, requiring a participating clinical peer before an AI-driven adverse determination, not on labeling or telling patients that AI was involved.
Who has to comply?
Health insurers, private review agents, and utilization-review entities operating in Georgia. If your coverage operation issues adverse determinations to patients in Georgia and uses AI in that process, this applies to you as of January 1, 2027.
Are other states doing the same thing?
Yes. SB 444 is part of a national human-in-the-loop wave on AI claim denials, with parallels in Arizona (HB 2175), Iowa (HF 2635), Indiana (HB 1271), and Washington (SB 5395). The specifics vary, but the shared principle is that a qualified person, not AI alone, must stand behind a coverage denial.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal, accounting, or audit advice. Confirm how any statute or requirement applies to your situation with qualified professionals in the relevant jurisdiction.