Trump admin moves to roll back AI transparency rules

Late last month, the Trump administration proposed sweeping changes to health information technology regulations that would eliminate requirements for software developers to disclose how their artificial intelligence tools are built and tested, a move that could reshape oversight of AI systems used in patient care.

The proposal, released by the Office of the National Coordinator for Health Information Technology, would roll back Biden-era requirements that health IT vendors submit detailed “model cards” describing their AI systems. These documents, akin to nutrition labels, outline how algorithms are developed, what data they use, and what risks they might pose to patients.

The changes are part of a broader deregulatory push outlined in the Health Data, Technology, and Interoperability: ASTP/ONC Deregulatory Actions to Unleash Prosperity proposed rule, known as HTI-5. The measure directly responds to President Trump’s executive order on “Unleashing Prosperity through Deregulation” and his directive to maintain American leadership in artificial intelligence.

“The HTI-5 proposed rule delivers on President Trump’s directive to reduce regulatory burden and to enable American innovation through artificial intelligence,” said Dr. Tom Keane, Assistant Secretary for Technology Policy and National Coordinator for Health IT in a statement. “These proposals reflect a commonsense approach that removes redundant requirements on health IT developers, that better ensures seamless patient access to their information and that sets a foundation for AI-based data exchange.”

Stay ahead of cutting-edge AI in healthcare—engage with the innovators, operators, and researchers shaping what’s next at ViVE in Los Angeles. AI Zone @ViVE 2026.

Massive overhaul of Health IT Certification Program

The proposal significantly pivots from the Biden administration’s approach to AI oversight in health care, which emphasized transparency and detailed disclosure requirements. Under the Biden-era rules, software developers were required to provide extensive documentation about their AI models, including information about training data, validation methods, and potential biases or limitations.

By removing these requirements, the Trump administration argues it will free up resources for innovation while reducing compliance costs. The agency estimates the HTI-5 rule would save $1.53 billion in total, including $650 million over five years for health IT developers, providers, and other stakeholders.

The proposed rule goes beyond AI disclosure requirements. It would eliminate more than half of the certification criteria in ONC’s voluntary Health IT Certification Program, which sets standards for electronic health record systems and related technologies. The changes would save certified health IT developers an estimated 1.4 million compliance hours in the first year alone, an average of up to 4,000 hours per developer, according to the agency.

The certification program streamlining aims to refocus the program on standards-based application programming interfaces, particularly those using Fast Healthcare Interoperability Resources, or FHIR, a framework for exchanging electronic health information. The administration says this shift would better support AI-enabled interoperability solutions through modernized standards.

The proposed rule also includes revisions to information blocking regulations, which govern how health care organizations share patient data. Based on stakeholder feedback, ONC is proposing to revise or remove certain terms, conditions, and exceptions in the information blocking rules to address potential misuse or abuse. The agency argues these changes would strengthen its ability to enforce the regulations and ensure patients can access their health information.

“These proposals reflect a commonsense approach that removes redundant requirements on health IT developers, that better ensures seamless patient access to their information and that sets a foundation for AI-based data exchange,” Keane said.

Concerns about patient safety and AI transparency

Health systems and clinicians are increasingly using AI tools for tasks ranging from analyzing medical images to predicting patient deterioration to drafting clinical notes. Proponents of lighter regulation argue that excessive requirements could stifle innovation and slow the adoption of potentially beneficial technologies.

However, the elimination of model card requirements may concern patient safety advocates and researchers who have called for greater transparency in health care AI. Model cards provide critical information about how algorithms are trained, what populations they’ve been tested on, and what limitations they may have. Without these disclosures, critics argue, it becomes harder to assess whether AI tools are safe and effective for diverse patient populations.

The Biden administration had implemented the model card requirement as part of its broader effort to ensure AI systems used in health care are developed responsibly. The requirement was designed to help health care providers understand the capabilities and limitations of AI tools before implementing them in clinical settings.

The HTI-5 proposal aligns with Trump’s broader executive orders on deregulation, including EO 14192 on “Unleashing Prosperity through Deregulation” and EO 14267 on “Reducing Anti-Competitive Regulatory Barriers.” The administration has framed the changes as necessary to maintain American competitiveness in AI development while reducing unnecessary regulatory burden on businesses.

ONC is encouraging the health IT community to carefully review the full proposed rule to understand all provisions being considered. The rule is now on display at the Federal Register and will be open for public comment for 60 days following publication.

The agency is also withdrawing certain proposals from the HTI-2 proposed rule that have not yet been finalized, though it did not specify which provisions would be withdrawn.

The public comment period will provide an opportunity for health care providers, technology developers, patient advocates, and other stakeholders to weigh in on the proposed changes. The final rule will likely reflect modifications based on this feedback, though the administration’s deregulatory direction appears firmly established.

The outcome of this rulemaking could have lasting implications for how AI is developed, deployed, and overseen in American health care. As AI becomes increasingly embedded in clinical decision-making, the balance between encouraging innovation and ensuring patient safety remains a central tension in health technology policy.