PATIENT AI Act discussion draft (2026)
AI-generated text. This page was generated using artificial intelligence.
PATIENT AI Act
United States — federal
Legislative discussion draft
TAM26B08 MJ1 S.L.C.; reviewed copy has no assigned bill number
Discussion draft released October 29, 2026
Proposed duties depend on enactment; sponsor feedback requested by December 1, 2026
November 3, 2026 — complete 45-page draft and sponsor materials
Overview and status
Senate HELP Committee Chair Bill Cassidy released the PATIENT AI Act discussion draft on October 29, 2026, seeking stakeholder feedback by December 1. The proposal addresses health-data safeguards, medical-device review, health-information-technology certification and Medicare coverage and payment. The reviewed text is an unnumbered discussion draft, rather than an enacted law.[1][2]
Privacy, liability and public-sector work
Title I would direct HHS to issue guidance on applying HIPAA privacy and security safeguards to AI, including data minimization, authorization, de-identification, secondary uses, contracts and cybersecurity risks. It also proposes a nonprofit-convened public meeting and recommendations on allocating liability among providers, facilities, insurers and developers, rather than directly establishing a new liability allocation. Other provisions cover a transparency study, HHS use cases, workforce training and research datasets limited to information already available for public use.[2]
FDA review and oversight
Title II would establish a voluntary review process for integrated devices, examining components and the safety of the complete device together. It proposes risk-based review of qualifying low- to moderate-risk software devices and predetermined change-control plans for software changes during pending reviews. Changes within an accepted plan would not, solely because of the change, require a new submission or restart the review period.[2]
Accredited independent third parties could review eligible software devices and recommend FDA approval or clearance. FDA would retain the ability to require additional review, while providing reasons. The draft includes accreditation, conflict-of-interest, audit and withdrawal safeguards, and a voluntary postmarket-surveillance pilot. Section 205 preserves existing federal and state liability, including product-liability and malpractice law, and the application of the existing federal device-preemption provision.[2]
Voluntary health-IT certification
Title III would expand clinical-research coverage in health-IT definitions and provide voluntary ONC certification for AI applications that are health IT but are not medical devices. Under section 303, incorporating AI or receiving this certification would not, by itself, turn otherwise excluded software into an FDA-regulated device.[2]
Medicare coverage, coding and payment
Title IV proposes a coverage-evaluation framework, a dedicated pathway for services using applications reviewed by certified entities, alternative payment methodologies and a process for AI-related HCPCS Level II codes. Existing Medicare benefit-category, coverage, provider and other requirements would remain. FDA clearance, third-party review or a billing code alone would not establish Medicare coverage or payment entitlement.[2]
The draft also proposes a five-year CMS Innovation Center model for selected autonomous or third-party-reviewed AI services, with beneficiary notice, clinical accountability, safety monitoring, data protections and preserved appeal rights. Its deadlines run from hypothetical enactment; publication of the draft does not activate them.[2]
Related coverage
References
- ↑ Senate HELP Committee, October 29, 2026 announcement.
- ↑ 2.0 2.1 2.2 2.3 2.4 2.5 2.6 PATIENT AI Act discussion draft, TAM26B08 MJ1 S.L.C., complete 45-page text, sections 1–405, reviewed November 3, 2026.