HHS's ASTP/ONC has proposed the HTI-5 rule, which would remove the certification requirement that clinical decision support and predictive AI developers publish model cards disclosing data sources and model attributes. Nelson Mullins notes these represented some of the only existing federal transparency guardrails for healthcare AI. If the rule is finalized, a health system can no longer assume a vendor's transparency documentation will be federally required. The due-diligence burden shifts inward: whatever a governance committee needs to know about a model's training data, intended population, and known limitations, it will have to ask for and verify itself. Worth tracking through the comment period.
Regulatory
A proposed federal rule would remove the model card requirement for clinical AI
A proposed HHS rule (HTI-5) would drop the model-card transparency mandate for clinical AI, shifting the due-diligence burden onto health systems.
CHARGE and Grid Health
CHARGE publishes independently. Grid Health, built in response to its members, runs the AI operations center health systems use to monitor models, detect drift, and govern AI in production.
Go deeper with Grid