AI-Native Healthcare: Navigating Compliance and Innovation

Since June 2026, Washington State has required a licensed physician or health professional to make any prior authorization denial based on medical necessity, and AI cannot be the sole basis (1). That rule captures what healthcare compliance AI tools must prove: who made the call, and on what evidence. Federal agencies relaxed their requirements in 2026 while states wrote new ones, so AI innovation in healthcare now runs on products that carry their own proof.
Key Takeaways
- Federal rules on healthcare AI eased in 2026, while 11 states passed new laws, including rules on insurer denials and therapy chatbots.
- Healthcare compliance AI tools should show, for any output, which inputs shaped it and who signed off.
- AI governance belongs to the health system that deploys the tool, and the Joint Commission started certifying that governance in June 2026.
- The strongest first projects are administrative workflows where AI prepares the work and a licensed person makes the decision.
How Is AI Regulation in the Healthcare Sector Changing?
AI regulation in the healthcare sector got lighter at the federal level in 2026 and heavier in state capitals. By late July, legislatures had passed more than a dozen new healthcare AI laws across 11 states (1).
On the federal side, the FDA revised its clinical decision support guidance in January. It now applies enforcement discretion to tools that give a single recommendation when only one is clinically appropriate, provided they meet its other criteria (2).
In October, ONC finalized the HTI-5 rule, which drops the requirement that certified health IT support access to 31 source attributes for predictive decision support, including intended use, training data and validation (3). ONC also said it may restore some of those attributes in future rulemaking, so stripping them out now would be a short-lived saving. The HIPAA Security Rule overhaul, proposed in January 2025, has slipped to a projected July 2027 (4).
The states went the other way. Iowa now requires clinicians to tell patients out loud before recording a visit for AI transcription, and Colorado banned AI therapy chatbots (1). A December 2025 executive order set up a Justice Department task force to challenge state AI laws (5). It cannot repeal them, though, and they apply until a court or Congress says otherwise.
For healthcare tech teams, this means fewer federal forms and more questions from buyers. When federal rules ease, the burden of proof shifts to the product and to the health system that buys it.
What Should Healthcare Compliance AI Tools Prove Before Go-Live?
Healthcare compliance AI tools should be able to answer four questions about any output before a health system lets them near patients. Each one maps to a rule already in force or to something buyers already ask in procurement.
- Which inputs shaped it? Since May 1, 2025, the Section 1557 rule has required covered entities to make reasonable efforts to identify decision support tools that use race, color, national origin, sex, age or disability as inputs, and to reduce the risk of discrimination (6).
- Who made the final call? Washington's denial rule is one example. Alabama's SB 63, effective October 1, 2026, also puts a licensed physician or health professional behind insurer denials (1).
- Did the patient know? Iowa's recording law makes disclosure part of the visit itself.
- How was it validated? HTI-5 removed validation details from EHR certification, but 88% of physicians in the AMA's 2026 survey called safety and efficacy validation critical to wider adoption (7).
You cannot piece any of this together from chat logs after an incident. It has to be captured while the tool runs. A tool that cannot show who approved an output is not ready for a clinical setting, no matter what its accuracy scores say.
Who Owns AI Governance in Healthcare?
AI governance in healthcare belongs to the organization that deploys the tool, and since June 2026, that organization can have it certified. The Joint Commission's voluntary Responsible Use of AI in Healthcare certification reviews a hospital's governance, safeguards, monitoring and staff education, and it does not certify AI products (8).
The certification builds on guidance the Joint Commission and the Coalition for Health AI published in September 2025 (9). Its seven elements cover governance structures, patient privacy and transparency, data security, ongoing quality monitoring, voluntary reporting of AI safety events, risk and bias assessment, and education and training.
For a healthcare tech vendor, each of those elements becomes a procurement question. That changes who the real buyer is. A tool can impress the CIO and still stall in front of a committee asking how bias was assessed.
Clinicians want a say, too, and 85% of physicians told the AMA they want to be consulted or directly involved in AI adoption decisions (7). A product that hands the governance committee its evidence ready-made spends less time stuck in review.
Where Is AI Innovation in Healthcare Paying Off?
AI innovation in healthcare tends to pay off first in the administrative work around care, where a person still signs off and an error can be caught before it reaches a patient. Adoption itself is settled. In 2026, 81% of physicians reported using AI in their practice, up from 38% in 2023 (7), and the FDA's list of authorized AI-enabled devices passed 1,600 by September (10).
Claims review shows why. In a claims review system Soluntech built, claims and supporting documents that used to sit in PDFs were structured so that each case shows its expected cost and potential risk before anyone acts on it. Low-impact cases get set aside before a reviewer spends time on them. The reviewer still decides, but now on cases where judgment changes the outcome.
That design lines up with where state law is heading. Washington and Alabama want a licensed person behind each denial, and a tool that prepares decisions instead of making them fits that rule from day one. The safest place to start is a workflow where AI prepares the decision and a licensed person still makes it.
How Do You Build Compliance Into an AI-Native Healthcare Product?
You build it in by settling the compliance questions before anyone picks a model. A healthcare product earns the AI-native label when the evidence a review board needs comes out of normal use, without anyone assembling it by hand after the fact. Six decisions carry most of the weight:
- Map where protected health information goes. Any vendor that creates, receives, maintains or transmits it for you, model hosts included, needs a business associate agreement.
- Record sensitive inputs. Log when a model uses any characteristic Section 1557 names, so the bias review has something concrete to check.
- Put sign-off where the law puts it. Route denials and medical-necessity calls to a licensed person, who can override the output with a recorded reason.
- Show disclosures inside the workflow. Put recording consent and AI-authorship notices for patient messages on the screen the clinician and patient use.
- Log the full decision path. Store the input, model version, output and the reviewer's action, which also supports HIPAA's audit controls standard (11).
- Watch the system after launch. Drift and override rates feed the ongoing quality monitoring the Joint Commission guidance expects.
The same records that make healthcare compliance AI tools auditable also make them easier to buy. Teams selling into Europe face a separate clock under the EU AI Act, which we cover in our guide to designing AI-native digital products for regulated environments.
Soluntech's work on healthcare software built around clinical workflows treats privacy and auditability as early architectural decisions. Adding these controls after launch costs more because, by then, the data model already limits what you can prove.
If you are deciding where AI belongs in a clinical or administrative workflow, start with the decision it will touch. Start the Validation Assessment to pin down the problem, the assumptions behind it and the next decision before you commit a build budget.
Sources
(1) Transparency Coalition, "States have passed 14 new laws this year regulating the use of AI in health care," July 27, 2026. transparencycoalition.ai/news/state-lawmakers-have-passed-15-new-laws-regulating-the-use-of-ai-in-health-care
(2) US Food and Drug Administration, "Clinical Decision Support Software," revised final guidance, January 6, 2026. fda.gov
(3) HealthSystemCIO, "HTI-5 Final Rule Drops Source-Attribute Requirements for AI and Other Decision Support Tools From EHR Certification," October 26, 2026. healthsystemcio.com/2026/10/26/hti-5-final-rule-ai-source-attributes/
(4) Fierce Healthcare, "Feds push back HIPAA security rule overhaul to July 2027." fiercehealthcare.com/health-tech/feds-push-back-hipaa-security-rule-overhaul-july-2027
(5) The White House, "Ensuring a National Policy Framework for Artificial Intelligence," executive order, December 2025. whitehouse.gov/presidential-actions/2025/12/eliminating-state-law-obstruction-of-national-artificial-intelligence-policy/
(6) 45 CFR 92.210, Nondiscrimination in the use of patient care decision support tools, final rule at 89 FR 37522, May 6, 2024. ecfr.gov/current/title-45/section-92.210
(7) American Medical Association, "AMA: AI usage among doctors doubles as confidence in technology grows," press release, March 12, 2026. ama-assn.org/press-center/ama-press-releases/ama-ai-usage-among-doctors-doubles-confidence-technology-grows
(8) Becker's Hospital Review, "Joint Commission launches certification for responsible healthcare AI use," June 1, 2026. beckershospitalreview.com/quality/joint-commission-launches-certification-for-responsible-healthcare-ai-use/
(9) Joint Commission and Coalition for Health AI, "Guidance on the Responsible Use of AI in Healthcare," September 17, 2025. chai.org/news/joint-commission-and-coalition-for-health-ai-chai-release-initial-guidance
(10) US Food and Drug Administration, "Artificial Intelligence-Enabled Medical Devices," content current as of September 22, 2026. fda.gov/medical-devices/digital-health-center-excellence/artificial-intelligence-enabled-medical-devices
(11) 45 CFR 164.312(b) and 164.514(b), HIPAA Security Rule audit controls and Privacy Rule de-identification standard. ecfr.gov/current/title-45/part-164
Soluntech Team
AI-Native Engineering Firm
Soluntech's team writes about software, AI systems, validation, and operational technology from the perspective of disciplined engineering and long-term system ownership.
Continue with related insights.
Explore more thinking connected to this topic, from software decisions to AI systems and operational execution.

Five Questions to Ask Before Going Live
The technical evidence worth examining before a working application becomes something people and businesses depend on. A working application can create a...
Soluntech Team
AI-Native Engineering Firm

What is app auditing, and how to know if your solution needs one
Imagine that you have developed an application for your business. It doesn't matter if you have done it from scratch or used a low- or no-code platform.
Soluntech Team
AI-Native Engineering Firm

The Real Reason 81% of Health System Leaders Are Frustrated With Their Own Organizations
McKinsey's latest survey reveals 81% of health system C-suite leaders say their operating model isn't working. The problem isn't technology — it's how decisions are made. Here's what it actually takes to build systems that perform.
Soluntech Team
AI-Native Engineering Firm