A wave of 2026 state laws lets AI assist in healthcare but bars it from making the consequential call alone. Indiana's and Tennessee's took effect July 1, and the principle reaches well past healthcare.
Congress still has not passed a real AI healthcare law. The states did not wait. This year a wave of them enacted rules that share one spine, and on July 1 the Indiana and Tennessee versions took effect: AI can assist, but a licensed human has to own the decision that reaches a patient.
The laws, read plainly
Per a Holland & Knight review of 2026 legislation:
| State | Law | What it does | Status |
| Indiana | HB 1271 | Bars health insurers from using AI as the sole basis to downcode a claim without a professional's review of the record. Also bars providers from using AI to submit claims without review. | Effective July 1, 2026 |
| Tennessee | SB 1580 | Bars AI systems and chatbots from advertising that they are qualified to act as licensed mental or behavioral health professionals. Enforced as a deceptive practice. | Effective July 1, 2026 |
| Delaware | HB 191 | Bars any nonhuman entity, including an AI agent, from being licensed as a nurse, physician, or physician assistant, or from using those protected titles. | Enacted April 2026 |
| Washington | SB 5395 | Bars health plans from relying solely on AI to deny, delay, or limit care. Only a licensed professional can make the adverse call, and plans must report AI-aided denials. | Enacted March 2026 |
Holland & Knight reads the full set the way I would. On the insurance side, "a health insurer cannot rely upon it as the sole basis for denying care." On the provider and chatbot side, a machine cannot hold a professional license, and it cannot claim to.
Why a law firm should read a healthcare bill
If you run a payer or a provider, this is your compliance calendar. If you run a law firm, an accounting practice, or a financial advisory, read it anyway. The principle these laws encode is the one every licensed profession is walking toward: a named, qualified human is accountable for the consequential decision, and the AI is support staff.
The enforcement is already attached. Tennessee now treats a false "licensed professional" claim as a deceptive act. Washington's law makes plans count their AI-aided denials. Accountability comes with a paper trail and someone empowered to check it.
This is the mantra, now in statute
This is Ayvede's view, and legislatures are printing it. AI drafts. You verify. The draft can come from a model. The signature has to come from a person who holds the license and can defend the call.
So the operational work is concrete. Name the human who reviews each AI-assisted output in a regulated workflow. Log that the review happened. Keep the model off the final approval. If a vendor's product is built so the AI issues the decision and a human "monitors," that design fails a plain reading of laws like Indiana's and Washington's. Fix the design before an auditor or a plaintiff reads it for you.
None of this bans the tool. Every one of these laws lets AI assist, triage, draft, and speed the tedious parts. They draw the line at the same place. The moment a decision lands on a patient, a person with a license stands behind it.
The takeaway: in regulated work, "a human in the loop" is becoming "a licensed human on the signature." Build for that while it is still a design choice and not an audit finding.
Ayvede helps high-trust firms put AI to work without handing it the decisions it cannot legally own. If your workflows need a qualified human accountable at the right step, that is the review we offer.
AI, Properly is the Ayvede field report.
Advise. Vet. Deliver.
Ask a question in writing