When AI art has no author: Generated images can't trace to training data (study)

MIT researchers discovered that in large-scale AI image generators, removing individual training images often doesn't change the output, a phenomenon they call attribution decay. This challenges the ability to trace generated images back to specific training data, impacting legal and ethical discussions about AI art authorship.