The CEO of America’s largest public hospital system has declared readiness to replace radiologists with AI for routine imaging reads like mammograms once regulations permit, igniting fierce backlash from physicians who accuse Nvidia’s Jensen Huang and Anthropic’s Dario Amodei of peddling “fiction” and “ignorance” to hype AI capabilities. Mitchell Katz, MD, president and CEO of NYC Health + Hospitals serving 1.1 million New Yorkers annually, argued during a Crain’s New York Business panel that AI could flag abnormalities while sidelining radiologists for “major savings,” particularly expanding breast cancer screening access.
Katz questioned fellow hospital CEOs on regulatory barriers preventing AI from reading images “without a radiologist,” citing current mammogram/X-ray interpretation as ripe for automation. His position echoes Anthropic CEO Dario Amodei’s recent podcast claim that AI has “taken over” radiology’s core functions, freeing physicians for human interaction, and Nvidia CEO Jensen Huang’s US-Saudi forum assertion that radiology has “largely converted to AI-driven” workflows with hiring actually increasing.
Radiologists Fire Back: Hype vs Reality
Neuroradiologist Ben White unleashed scathing rebuttals, calling Huang’s claims “sheer unadulterated fiction.” “AI simply doesn’t drive a meaningful part of radiology workflow,” White wrote. “Some triage lists and algorithms detecting blood/fractures haven’t changed the game. Generative AI drafting impressions represents the only limited efficiency gain.”
Dismissing Amodei as ignorant—or worse, deliberately misleading—White questioned whether AI executives peddle radiology fables for fundraising: “Dario Amodei is worth $7 billion… with Anthropic raising at $380 billion valuation. Maybe they know it’s not true but need the storytelling.” Huang’s hiring surge claim ignores AI’s minimal workflow penetration despite vendor promises.
AI’s Real Role vs Executive Narratives
Katz’s proposal targets high-volume, low-variability screening where AI sensitivity exceeds human consistency, potentially scaling access in underserved areas. FDA-cleared tools like GE Healthcare’s CRAIO (94% sensitivity) and iCAD’s ProFound AI (8.4% miss reduction) demonstrate triage value, but no system matches radiologist specificity for complex cases requiring clinical correlation.
The controversy exposes a pattern: AI CEOs repeatedly nominate radiology as first-to-be-replaced despite contradictory evidence—RSNA 2025 data shows 7% radiologist growth vs 3% physician average. Workforce augmentation, not replacement, defines reality: 82% of practices use AI per ACR, prioritising triage over elimination.
Regulatory realism tempers Katz’s vision: New York’s strict licensure demands human oversight; FDA 510(k) clearances limit standalone diagnostics. While NYC Health explores pilot programs, full replacement faces legal, liability and union barriers.
The debate reveals AI’s dual reality: genuine screening augmentation meets executive overreach framing incremental tools as existential threats. Radiologists demand evidence over narrative as hospitals weigh cost pressures against diagnostic integrity.
