When AI art has no author: Study finds generated images often can’t be traced to training data

Who gets credit when an AI generates an image? It's a central question in copyright lawsuits against AI companies worldwide. New research from MIT finds that the larger the training dataset, the less any single image influences the final output. Using a new testing method, researchers proved that removing an artist's entire body of work—or any single image—from the training data doesn't change what the model produces. If removing something changes nothing, the researchers argue, it can't have influenced the output. The finding challenges copyright cases built on attribution and raises questions about whether AI-generated images are derivative works or entirely new creations.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton