In the past year, tech giants like OpenAI, Suno, Udio, and Stability AI have leaned on “fair use” to train models with copyrighted material—generating billions, with OpenAI alone reaching over $3 billion in revenue in 2024. But with lawsuits looming, the future of AI rests on one thorny legal question: What exactly is fair use?🧐

Here are 4 key questions to consider when evaluating whether a use is “fair”:

1. Purpose & CharacterIs it for commercial gain or nonprofit educational use? Has enough new meaning or purpose been added to transform the original work?

2️. Nature of the Original WorkWhat type of work is being used? Creative works (music, film) enjoy stronger copyright protections than factual works (like data compilations).

3️. Amount & Substantiality of the Portion UsedHow much content is being used, and how integral is it? A snippet or the whole piece?

4️. Effect on Market ValueCould the new use harm the market for the original work? The new work shouldn’t serve as a replacement.⚖️ Fair use can be frustratingly vague—it’s not about fixed rules but a sliding scale of these four factors.

As AI use grows, we expect to see more companies asked to license the data used for training.👉 What do you think? Will fair use evolve to meet AI’s needs, or will licensing become the new norm?