Doctors Not AI Act

By: David Ostrowsky

While many healthcare organizations have lauded AI for its inherent potential to improve the US healthcare system by expediting claim review decisions, generating summaries of doctors’ conversation with patients, and analyzing scans and X-rays among other functions, there has recently been strong pushback on Capitol Hill. So strong, in fact, that on September 1, a new bipartisan bill that would prevent employer-sponsored health plans, including self-insured ones, from using AI technology to deny health insurance claims and prior authorization requests was introduced in the U.S. House of Representatives.

At its core, the Doctors Not AI Act (H.R. 10210), spearheaded by Rep. Greg Landsman, D-Ohio and under the jurisdiction of the House Energy & Commerce Committee, the House Ways & Means Committee, and the House Education & Workforce Committee, would mandate that a licensed healthcare professional monitor AI systems utilized to review medical claims and be held accountable for any “adverse benefit determinations.” In other words, the claim procedure requirements would dictate that adverse benefit determinations involving clinical judgment—whether upon initial claim review or appeal—could not be issued if they were ultimately determined by an AI system.

Per the text of the bill, denials could be “made only by a licensed health care professional who is acting within the scope of the professional’s license and who has training and experience in the provision of the health care item or service that is the subject of the determination.” While health care professionals are still permitted to deploy AI technology, they would need to disclose such usage (many states already require disclosure of the clinical criteria used to deny a given claim and whether a human clinician reviewed it) and retain records pertaining to any AI system outputs, scores, recommendations or determinations. It’s also worth noting that the Doctors Not AI Act has specific protections for mental health and substance use disorder prior authorization requests—a realm of healthcare where some feel that AI-fueled denials have been especially prominent.

The Doctors Not AI Act, which if implemented as written could take effect on January 1st of the first calendar year starting at least 12 months after the date of enactment, would also incorporate a definition of “artificial intelligence system” into section 2791 of the federal Public Health Service Act as well as insert an AI system definition into section 3 of ERISA and AI claim procedure requirements into ERISA section 503. While the ERISA modifications would apply to self-funded plans, the Doctors Not AI Act would apply the requirements to fully insured group health plans by making similar adjustments to Internal Revenue Code section 9832 and IRC section 9815.

Irrespective of whether the Doctors Not AI Act becomes implemented as written, AI projects to continue having a rigorous foothold in healthcare. Over the past decade, AI technology has transformed how diseases are detected, how medical data is evaluated, and, ultimately, how physicians make decisions. In this regard, AI can represent an endlessly valuable tool in the medical field. But, as most would concur, it cannot serve as the ultimate arbiter for medical decisions.

Thus, with AI and healthcare being forevermore intertwined, the conversation has shifted to how the medical community can responsibly deploy the technology for patients’ well-being. And it’s not exactly a simple conversation. In an absolute worst-case scenario from a clinician’s perspective, AI, while examining an X-ray of a patient, may not catch a malignant tumor; the subsequent incorrect reassurance that everything is fine could deter the patient from seeking treatment as the otherwise curable cancer spreads throughout their body. And then there’s the widespread concern among many patients that when it comes to AI-fueled insurance practices, AI has potential to rubber-stamp denials at a breakneck pace. Whereas only a decade ago, prior authorization (PA) requests for procedures and surgeries were reviewed by actual human clinicians who grounded their decisions in medical evidence, it has become increasingly common for PA reviews to be executed by algorithms that—in some people’s opinions—safeguard the insurers’ best interests. In fact, according to the American Medical Association 2026 Prior Authorization Physician Survey, 61% of physicians indicated that AI tools are increasing the number of prior authorization denials they face while 29% reported serious adverse patient outcomes stemming from PA delays. Of course, many of these patients whose PA requests repeatedly get rejected are those who lack the financial means and other resources to properly appeal the decision.

While the Doctors Not AI Act is still in its nascent stages of development and faces considerable headwinds, it’s undeniable that there is federal momentum for reform (on both sides of the aisle) while some states have taken matters into their own hands. If nothing else, this new bill reflects millions of Americans’ collective frustration with struggling to obtain adequate healthcare access—especially when a real live human is not behind the ultimate decision.