MIT study: AI-generated images often can't be traced to training data

Warm_Ad1257 · reddit · 2026-08-20

A new study from MIT finds that AI-generated images often cannot be traced back to the specific data used to train them. The research explores the data forgetting and anonymization properties of generative models, showing that generated images largely cannot be used to reverse-engineer original images from the training set, sparking further discussion on authorship and data copyright in AI art.

Related event: MIT CSAIL Reveals "Attribution Decay": AI-Generated Images Can Rarely Be Traced to Training Data(7 posts)→

Original post →

More from Safety

Safety channel →