Is AI Training Fair Use?

Fair use is not a single switch. It is a balancing test, and AI training cases are testing its boundaries.

Factor one: purpose and character

AI companies argue that training is transformative because models learn statistical relationships rather than republishing books or images. Plaintiffs argue the copying is commercial and competes with licensing markets.

Factor two: nature of the work

Creative works often receive stronger protection than factual works. Large datasets usually mix both.

Factor three: amount used

Training often uses entire works. That can weigh against fair use, though courts sometimes allow complete copying when it is necessary for a transformative purpose.

Factor four: market effect

This may become the most important factor. Courts will ask whether training harms existing or plausible licensing markets, or whether generated outputs substitute for original works.