Overview
- A peer-reviewed Radiology study tested 264 X-rays with 17 radiologists from six countries and found synthetic images often passed as real.
- Doctors detected fakes only 41% when unaware, improving to about 75% after being told that synthetic images were included.
- Large AI models from OpenAI, Google, and Meta showed uneven detection, with accuracy ranging from roughly 52% to 89%.
- Researchers cataloged recurring tells such as overly smooth bones, unnaturally straight spines, and uniform vessel patterns, and released a training dataset with quizzes.
- To curb misuse in courts and hospitals, the authors call for image provenance at capture using invisible watermarks or cryptographic signatures and warn synthetic CT and MRI are likely next.