STAT Reporter Matches Radiologists in Spotting Deepfake X-rays, Raising AI Patient Safety Alarms

A non-expert journalist has matched the performance of trained radiologists in identifying AI-generated deepfake X-rays, exposing a critical vulnerability in medical imaging. In a recent study published in *Radiology*, 17 radiologists correctly differentiated real from synthetic images only about 75% of the time. STAT reporter Katie Palmer, who reported on the findings, took the same test and scored an identical 75%, demonstrating that the AI forgeries are sophisticated enough to challenge both professional and untrained eyes.

The study's results signal a direct threat to diagnostic integrity and patient safety. The ability of generative AI to create convincing, yet fabricated, medical images introduces a new vector for error or potential fraud within clinical workflows. In a follow-up discussion on the STATus Report podcast, host Alex Hogan explored these implications with Palmer and took the quiz himself, testing whether he could surpass the scores of both the radiologists and his colleague.

This development places immediate pressure on radiology departments, medical software vendors, and regulatory bodies. The core challenge is no longer a distant theoretical risk but a present capability that could undermine trust in diagnostic evidence. The medical community now faces the urgent task of developing and deploying reliable detection tools and updated verification protocols before these AI-generated images potentially infiltrate patient records and influence treatment decisions.