AI Copyright Issues: OpenAI’s Use of YouTube Content

This week, when asked about the training data for OpenAI’s Sora, CTO Mira Murati mentioned it included licensed or ‘public’ content, without specifically addressing the use of YouTube videos. This ambiguity raises critical questions about the legality of using ‘publicly available’ content in AI training, amidst the intricate landscape of copyright law.

This distinction is pivotal, as ‘publicly available’ does not inherently mean free from copyright claims, challenging the assumption that such content can be used freely for training AI.

Debate online on this topic has been intense.

If you have not seem it, watch from 4:20 here (not kidding):

The issue is complex. In the US, the owner of copyrights or personality rights is generally the only person who can monetize them. They can permit others to monetize those rights.

A key exception in US copyright law, ‘fair use,’ allows for limited use of copyrighted material without permission for purposes like criticism, comment, news reporting, teaching…