A note from Re:Create: Innovation Isn’t Stealing

Anti-AI crusader (and talented actor, TBH) Joseph Gordon-Levitt launched a splashy campaign this week with some big name celebrity endorsements for a petition proclaiming “Stealing isn’t innovation.” Ironically, the sign-on campaign is itself an obvious remix, an American reboot of British campaigner Ed Newton-Rex’s equally strident “statement on AI training” from 2024. I’m sure Joe got a license. (Read our take on Ed’s self-defeating campaign, which applies equally to Joe’s.) In any case, both efforts are deeply mistaken, not only about copyright, but about creativity.

Fundamentally, of course, copying isn’t theft. As cartoonist Nina Paley whimsically explains, “Stealing a thing leaves one less left; copying makes one thing more.” It may seem semantic, but it’s important, because we’ve never treated all copying as theft. Joe’s petition looks a lot like Ed’s petition, Looper looks a lot like 12 Monkeys, Star Wars looks a lot like…Dam Busters (seriously), Ed Sheeran songs sometimes use the same chord progressions as Marving Gaye songs, and, well, nothing comes from nothing. Everything is a remix. Creativity always involves copying. If copying were theft, and no one could ever use something that came before as an ingredient or an inspiration for a new creation without asking a rightsholder, creativity would be impossible for all but the wealthiest entertainment conglomerates. 

Artists tend to understand this gift-like nature of creativity intuitively, and I’m sure the signatories of this petition wouldn’t disagree that “taking” from the art you love in order to make more art is not the same as stealing. They’d probably disagree that AI training is a creative endeavor, but their statement is much blunter than that: they say AI training is “stealing” because it copies without a license. When art is your business, and new technology seems to threaten that business, it is understandably hard to hold on to first principles, or to apply them to creativity that looks different from your own. That explains why these folks are taking a maximalist position on the scope of copyright, but it doesn’t excuse it.

Embracing a radically proprietary, commodified view of art (which is what this campaign represents: a demand that every productive encounter with a creative work, every reuse of its ideas, parts, and pieces, should be metered and paid-for) will not save creative workers from AI disruption. Most commercially significant copyrights are not owned by creative workers. Actors, in particular, almost never hold copyrights in their work. In the eyes of the law, if training on movies is “theft” from anyone, it’s theft from the studios who commissioned their performances. Insisting on copyright licenses as a precursor to AI training won’t give most actors, musicians, or other creative workers any more control, or any more money, than they currently have. It will not stop the use of AI technology in moviemaking or other artforms, nor will it prevent the impact of that use on workers. Nor will it have any appreciable impact on “Big Tech” companies, who can afford to pay for licenses. It will just give the handful of oligopolies who control most copyrights a chance to cash in on AI and potentially warp it to protect their own interests. 

The people who are already being harmed by legal threats and heated rhetoric about theft are non-profits, scholars, start-ups, and their fellow creators who are developing their own novel AI tools. Someone needs to tell Joe that their innovation isn’t stealing.