AI-Generated Deepfake X-Rays Deceive Radiologists and AI Systems

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

A multi-center study found that radiologists and advanced AI models cannot reliably distinguish AI-generated deepfake X-ray images from authentic ones. This vulnerability exposes healthcare to risks such as misdiagnosis, fraudulent litigation, and cybersecurity threats, highlighting the urgent need for improved detection tools and training.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves AI systems explicitly (ChatGPT, RoentGen, and other LLMs) generating synthetic medical images that are indistinguishable from real ones by experts and AI detectors. This misuse or potential malicious use of AI-generated deepfakes can directly lead to harms such as fraudulent litigation, clinical misdiagnosis, and cybersecurity attacks causing clinical chaos, all of which are harms to health and communities. The study documents these risks with evidence of actual AI-generated images deceiving professionals, thus meeting the criteria for an AI Incident rather than a mere hazard or complementary information.[AI generated]
AI principles
SafetyRobustness & digital security

Industries
Healthcare, drugs, and biotechnologyDigital security

Affected stakeholders
ConsumersWorkers

Harm types
Physical (injury)Economic/Property

AI system task:
Content generation


Articles about this incident or hazard