Senate's PATIENT AI Act Would Let AI Expand in Healthcare — With Guardrails

Sen. Bill Cassidy introduced the PATIENT AI Act on November 2, legislation that would expand AI's role in healthcare while establishing the first comprehensive federal patient safety framework for diagnostic AI. The bill lands as a new survey shows nine in ten patients want explicit consent required before AI is used in their diagnosis — and nearly two-thirds of healthcare providers are actively planning to deploy AI for claims and denials management, a high-stakes use case almost entirely outside current FDA oversight.

The PATIENT AI Act — its full name has not been publicly detailed in available bill text — is Cassidy's attempt to move past years of congressional debate about healthcare AI into an actual legislative framework. The bill is designed to give both regulatory predictability to AI developers and enforceable patient safety standards to the medical system. Its arrival in the Senate marks a turning point: sector-specific AI legislation, rather than broad horizontal AI law, may be the path Congress actually takes.

The consent gap the bill is trying to close

The survey data released alongside the bill tells a story about mismatched expectations. According to Healthcare IT Today's reporting of the findings:

• 91% of patients want their explicit permission required when AI is used in their diagnosis
• 64% of healthcare providers plan to deploy AI specifically for claims and denials management
• The two numbers exist in tension: patients want consent and transparency; providers are deploying AI primarily in administrative and financial roles where patient-facing consent mechanisms are not standard practice

Most AI diagnostic tools currently deployed in U.S. hospitals do not require or display patient consent prompts. A physician may use an AI system to flag a suspicious finding in a chest CT scan, and the patient may never know. Whether that constitutes a disclosure failure depends entirely on how you read existing informed consent doctrine — which was written before diagnostic AI existed.

The PATIENT AI Act's consent framework is one of the least-detailed publicly available provisions. Whether it mandates disclosure, requires opt-in consent, or establishes an opt-out structure will determine whether it is a minor administrative change for hospitals or a significant operational overhaul.

The claims and denials problem

The 64% figure on claims and denials AI deployment deserves separate attention, because this is where AI is already making consequential healthcare decisions — and where the stakes for patients are highest.

Prior authorization AI and claim denial systems determine whether a patient receives care their physician prescribed. An AI system that incorrectly denies a claim for chemotherapy, a post-surgical rehabilitation stay, or a specialty drug is not making an administrative error; it is making a clinical outcome decision, disguised as a billing decision. These systems are not FDA-cleared medical devices. They are revenue cycle tools. The regulatory gap is significant.

The PATIENT AI Act's scope — whether it covers back-office AI systems like denials tools, or only FDA-cleared diagnostic AI — is the most consequential design decision in the bill and was not specified in Cassidy's November 2 announcement.

What the bill would do

Based on available summary information, the PATIENT AI Act is expected to:

• Establish regulatory predictability for AI developers in healthcare — clarifying what FDA review processes apply to which categories of healthcare AI
• Create federal patient safety standards for AI used in diagnostic settings
• Provide a framework for patient consent or disclosure when AI participates in clinical decisions
• Address the oversight gap created by the proliferation of healthcare AI tools that do not currently require FDA clearance

The bill is in committee. No floor vote has been scheduled.

The political environment

Cassidy has positioned the PATIENT AI Act with bipartisan framing: regulatory predictability appeals to industry and Republican members; patient safety and consent appeal to Democratic members and consumer advocates. That framing may give this bill more runway than single-issue healthcare AI bills that preceded it.

The AMA and major hospital associations have not yet publicly responded to the November 2 introduction. Their position on the consent provisions — particularly any requirement that applies to deployed AI tools rather than only new ones — will determine whether the bill finds institutional support or resistance from the healthcare industry itself.

What to watch

Whether Cassidy attracts Democratic co-sponsors in the first 30 days is the clearest signal of the bill's viability. Also watch for CMS and FDA testimony when the bill reaches committee hearings — those agencies will have significant input on how the bill interacts with existing regulatory frameworks, and their posture will shape amendments. The claims and denials scope question, if clarified in markup, will be the most commercially significant detail in the bill's evolution.