HHS finalizes HTI-5 rule removing AI transparency requirements

The Department of Health and Human Services (HHS) has finalized the HTI-5 rule, which removes several certification criteria for electronic health records. A significant change includes the elimination of requirements for developers to disclose source attributes for predictive decision support interventions. Previously, certified health IT was required to provide access to specific data regarding a model's intended use, training data, validation methods, and performance.

While the rule removes 14 of 60 certification criteria, it leaves 39 unchanged and revises seven others. HHS estimates the move will result in $1.55 billion in regulatory savings. However, critics express concern that removing these transparency measures—such as intervention risk management and source attributes—compromises clinical decision support safety. Under the new rule, clinicians may need to manually request information from vendors that was previously provided automatically.

ONC has noted that while privacy and security criteria remain for now, future rulemaking may address them. The rule also scales back real-world testing requirements and reduces insights reporting, though it retains the FHIR bulk data access measure.