MIT study finds AI image attribution weakens as training data grows

DIGITIMESAugust 24, 2026en

MIT researchers said a larger training set makes it harder to trace an AI image back to a specific source, a pattern they call attribution decay. The findings, reported in connection with diffusion models, could complicate copyright disputes and regulatory scrutiny over generative AI.

This is a short summary published by AI Global Wire. The full article is owned and hosted by DIGITIMES — open it there to read it in full.

• Forskning
• Bild