AI vs. Radiologists: The Hospital CEO Who Just Fired the Starting Gun

A bombshell statement from one of America's most powerful hospital executives has ignited a fierce debate across the medical world. Mitchell H. Katz, MD, president and CEO of NYC Health and Hospitals, the largest public hospital system in the United States, stepped onto a panel stage on March 25, 2026, and declared that his organization could already replace a significant number of radiologists with artificial intelligence. The only thing standing in the way, he argued, is regulation.

The Man Behind the Statement

Mitchell Katz is not some tech entrepreneur with a flashy pitch deck. He is a trained internal medicine specialist who has led the 11-hospital NYC Health and Hospitals network since 2018. The system serves over one million New Yorkers annually, making it the largest public hospital network in the country. He was speaking at a panel discussion organized by Crain's New York Business. The debate about the rise of AI doctors and whether physicians trust them has been building for years, and Katz just turned it into a full-scale policy fight.

The $224 Million Man

Katz was not speaking hypothetically. He was speaking as a sitting CEO who oversees a system that has already invested $224 million in upgrading its imaging technology infrastructure. This is someone who knows the dollars, the data, and the direction his system is heading. He was not pitching a future vision. He was describing a present-day capability that is being held back by one thing only: the existing regulatory framework in New York state.

What Exactly Did He Say?

Katz was direct. "We could replace a great deal of radiologists with AI at this moment, if we are ready to do the regulatory challenge," he told the audience at the Crain's panel. His argument centered on a clear model: AI handles the first read on a medical image, and a human radiologist steps in only when the algorithm flags something abnormal. For normal results, particularly in lower-risk screening scenarios like mammograms, AI would take the wheel entirely.

The Cost-Cutting Motive Is Real

Let's be honest about what is driving this conversation. Radiologists are expensive, and their costs are rising. An aging population, expanded screening guidelines, and increasing imaging volumes have created serious budget pressure for hospitals. Katz argued that allowing AI to handle initial reads on routine, low-risk screenings could generate "major savings." A closer look at how AI is unlocking better outcomes in healthcare shows that cost reduction has consistently been one of the strongest arguments made for wider AI adoption across hospitals.

Other CEOs Are Already On Board

Katz was not alone on that stage. David Lubarsky, MD, MBA, president and CEO of the Westchester Medical Center Health Network, said his system is already deploying similar AI tools. He told the audience that the AI his network uses is "actually better than human beings" at detecting breast cancers. For women who are not high risk, the negative AI read carries an error rate of roughly three cases in every ten thousand. Sandra Scott, MD, CEO of One Brooklyn Health, called the prospect of AI-led image reads "a game-changer" for her tight-budget safety-net institution.

Radiologists Fire Back Hard

Not everyone in medicine is cheering. Mohammed Suhail, MD, a San Diego-based radiologist with North Coast Imaging, called Katz's comments "undeniable proof that confidently uninformed hospital administrators are a danger to patients." He said hospital CEOs are "easily duped by AI companies that are nowhere near capable of providing patient care." His conclusion was stark: "Any attempt to implement AI-only reads would immediately result in patient harm and death." He also noted that hospitals are willing to cut costs even at the expense of patient safety, as long as it remains legal.

The Regulatory Wall Blocking the Move

Here is the critical detail that often gets lost in the headlines. New York state law currently requires a licensed radiologist to sign off on any diagnostic image before it is finalized. Katz cannot simply flip a switch and deploy AI autonomously. He needs regulators to change the rules first. That is exactly what he was rallying fellow hospital CEOs around at the Crain's panel. The outcome of this regulatory debate could set a national precedent for how AI-led radiology is handled across the entire country.

FDA Has Already Opened the Door

It is worth noting just how deep AI has already penetrated the radiology space. Roughly three-quarters of the more than 1,000 AI applications cleared by the FDA for use in medicine are in radiology. These tools primarily help flag abnormalities, improve image quality, and automate routine tasks. The regulatory groundwork at the federal level is already laid. What Katz is now pushing for is a state-level change that would allow the technology to operate without a licensed radiologist serving as the mandatory final authority on every scan.

The Stanford "Mirage Effect" That Should Terrify Everyone

Just as this debate was heating up, troubling new research landed on the skeptics' side. In a yet-to-be-peer-reviewed Stanford University study, researchers found that AI chest X-ray tools can score impressively on medical benchmarks without ever seeing a real X-ray. Rather than admitting no image is present, these systems construct elaborate, confident explanations for findings on images they never accessed. As Futurism reported, the Stanford scientists called this "epistemic mimicry," where a model simulates perception while being anchored to no image at all.

When AI Diagnoses Diseases That Aren't There

The mirage effect goes beyond theory. According to The Decoder, frontier AI models confidently described visual details in over 60 percent of cases when given no image at all. With standard evaluation prompts, that rate climbed to between 90 and 100 percent. Fabricated diagnoses skewed toward severe conditions including heart attacks, melanomas, and carcinomas. This is precisely why platforms like OpenAI's ChatGPT Health have drawn intense scrutiny over readiness for high-stakes medical environments. A missed image upload could trigger life-altering treatment for a condition that does not exist.

Deepfake X-Rays Are Already Fooling Experts

The safety concerns do not stop at hallucinations. A study published in the journal Radiology found that neither experienced radiologists nor advanced AI models can reliably distinguish real X-rays from AI-generated deepfakes. Researchers tested 17 radiologists across six countries. Even when warned synthetic images were present, radiologists averaged only 75 percent accuracy in identifying them. When not warned at all, only 41 percent noticed anything unusual. Years of experience offered no advantage. Researchers warned this vulnerability could enable fraudulent injury claims or allow hackers to inject fake images into medical records.

Dario Amodei Stepped Into This Fight Too

Katz is not the first high-profile voice to claim AI is ready to take over radiology. Dario Amodei, PhD, CEO of Anthropic, made similar statements in a podcast interview, claiming AI had already taken over the core function of the specialty. Radiologists criticized those remarks sharply. Mohammed Suhail pushed back on Amodei's comments with the same intensity as his response to Katz. A prominent AI company CEO and a major hospital CEO making virtually identical claims in quick succession is not a coincidence. It signals a growing administrative and tech-industry consensus that the transition is closer than clinicians believe.

What a Radiologist's Job Actually Involves

Critics argue that hospital administrators are reducing a complex profession to a single task. A radiologist does not merely classify an image. The role also includes triaging cases, training medical residents, and signing diagnoses, none of which current AI systems can perform. Economist Luis Garicano, co-author of a recent report on the subject, pointed out that image classification is just one component of a much broader clinical role. As one expert stated plainly, "Autonomy requires zero hallucination, consistent findings, and a clear legal and regulatory framework, which we just don't have yet."

Jensen Huang's Warning Nobody Wants to Hear

Nvidia CEO Jensen Huang recently offered a sobering counterpoint to the replacement narrative. Speaking on Lex Fridman's podcast, Huang said earlier alarmism about AI replacing radiology discouraged people from entering the field, creating a specialist shortage even as AI became standard in every radiology platform. His view is that AI makes radiologists more capable, not redundant. "Because you are able to study scans so much faster now, you can study more scans. You can diagnose better. Hospitals are making more money. You need more radiologists," he said. Radiologist numbers and salaries have both continued rising despite AI adoption.

Where This Battle Goes From Here

The debate sparked by Katz's March 25, 2026 comments is far from over. It is just beginning. The regulatory battle over New York state imaging rules is live, and its outcome could reshape hospital AI policy nationwide. Administrators are watching budgets. Radiologists are organizing opposition. AI developers are refining tools. Regulators are navigating territory with no established playbook. AI shows genuine promise in expanding access and cutting costs for low-risk screenings. But the same technology has been shown to fabricate diagnoses and deceive experienced clinicians. The gun has been fired. The finish line keeps moving.

Source & AI Information: External links in this article are provided for informational reference to authoritative sources. This content was drafted with the assistance of Artificial Intelligence tools to ensure comprehensive coverage, and subsequently reviewed by a human editor prior to publication.