What We Told HHS About AI Governance in Healthcare

Why This RFI Matters

In early 2026, HHS published a Request for Information titled "Accelerating the Adoption and Use of Artificial Intelligence as Part of Clinical Care." The RFI asked stakeholders to identify governance challenges, recommend how HHS can support private sector frameworks, and describe who inside health systems actually controls AI adoption decisions.

When HHS asks how to accelerate AI in clinical care, the answers shape what federal support looks like: funding priorities, technical assistance programs, the expectations that accreditation bodies set.

The risk is that the conversation stays abstract. Many RFI responses will focus on what HHS should regulate, what vendors should disclose, or what the FDA should do differently. Those are valid questions, but they skip the problem most health systems are actually facing: they don't have the internal governance structures to manage AI tools they've already deployed.

Our response focused on three questions from the RFI, each addressing a different dimension of this governance gap.

The Governance Gap Is Real, and Measurable

The RFI asked about novel legal and implementation issues that challenge existing governance structures. Our answer: the most significant challenge isn't legal ambiguity at the federal level. It's organizational ambiguity within health systems themselves.

Over 70% of acute care hospitals now have AI tools influencing clinical decisions, whether through EHR-embedded algorithms, operational scheduling tools, or documentation assistants. Many fall outside FDA device regulation. "What federal rules apply?" often has a clear answer. "Who inside our organization is accountable for this AI-influenced decision?" usually doesn't.

The data tells a consistent story:

That last finding is worth sitting with. If the VA, one of the largest and most sophisticated healthcare organizations in the country, can deploy AI without basic safety coordination, the governance gap isn't a small-hospital problem. It's an everyone problem.

The Digital Divide in Governance Capacity

The governance gap doesn't affect all organizations equally. Our response highlighted a growing divide: 86% of system-affiliated hospitals use predictive AI versus just 37% of independent facilities. And among the 18% of health systems with mature AI governance, three-quarters have over $1 billion in net patient revenue. Governance maturity tracks closely with organizational size and resources.

This matters because federal policy is trending toward less prescriptive requirements. HTI-5's proposed changes would shift more responsibility for evaluating AI safety, bias, and fitness-for-purpose from federal certification to the organizations deploying these tools. Well-resourced health systems with dedicated informatics leadership can navigate that shift. Community hospitals and safety-net providers face a governance burden they don't have the capacity to meet.

The organizations that most need governance support don't have the resources to build it alone. That's the gap federal policy should be closing.

Supporting Private Sector Frameworks

The RFI asked how HHS can best support private sector activities to promote effective AI use. Our argument: the private sector has already started building governance infrastructure, and HHS should be supporting that work rather than duplicating it through regulation.

The most significant development is the Joint Commission and Coalition for Health AI guidance published in September 2025. Because the Joint Commission holds CMS deeming authority and accredits more than 80% of U.S. hospitals, health systems are treating this guidance as a roadmap for future accreditation expectations, even though it doesn't create new standards yet.

Professional societies are also building AI-specific infrastructure. The American College of Radiology has launched Assess-AI, a quality registry for monitoring imaging AI performance in real-world settings, and is working toward a formal AI accreditation program. The AMA's STEPS Forward toolkit provides practical governance guidance for health system leaders.

We recommended four specific ways HHS can support this work:

The risk of heavy federal regulation is that it freezes current practice into compliance requirements before the field has learned what actually works. The opportunity in supporting private sector frameworks is that governance can evolve with the technology while still establishing reasonable expectations.

Who Actually Controls AI Adoption?

The RFI asked which roles and decision-makers have the most influence on AI adoption. Our answer: no single role consistently owns AI decisions across health systems, and that fragmentation is itself a primary hurdle.

The result is a structural disconnect: the person who approves the investment (often the CFO) is rarely the person who understands what's being approved. Meanwhile, 33% of organizations cite "unclear ownership" between IT, clinical, and compliance teams as a primary barrier to governance.

Governance Accelerates Adoption

Here's what surprised people in our response: governance doesn't slow AI adoption. It actually accelerates it.

This makes sense when you think about what governance actually provides: a defined process for answering basic questions. Who approves this tool, who validates it against our patient population, who monitors it after deployment, and who decides when to turn it off? Without those answers, every AI decision becomes an ad hoc negotiation, and organizations stall.

The primary administrative hurdle isn't technology cost or complexity. It's that most organizations don't have a defined process for answering basic governance questions.

Jim Younkin, MBA, FACHDM

CTO & Co-Founder, Mosaic Life Tech. 30+ years in health IT including leadership roles at ONC/ASTP, founding Pennsylvania's first regional health information exchange serving 4M+ patients, and advising healthcare organizations on AI governance.