Copyright Law Could Still Cripple the Entire Generative AI Industry, New Ruling Shows

Illustration via MidjourneyPhoto byBay Area Telegraph

With AI companies’ gung-ho attitude about building their models–and the hundreds of billions of dollars they’re pouring into training–it’s easy to assume that the generative AI industry is built on solid foundations.

But as a recent ruling in an ongoing lawsuit shows, copyright law could still cripple the entire industry.

The lawsuit, filed by artists against Stability AI and others, accuses these companies of using copyrighted artistic works without permission to train their AI models, including Stable Diffusion, which then allegedly generates images that can mimic or closely resemble the original works.

Note: This article is for informational and reporting purposes only, and does not constitute legal advice.

The Ruling

In recent rulings, the U.S. District Court for the Northern District of California has allowed some of the artists’ claims to move forward, particularly those related to direct copyright infringement during the AI training process.

This indicates that the courts recognize the potential for AI models to infringe on copyrights through the use of large datasets of images scraped from the internet. It’s a major blow to generative AI companies, and a huge win for the artists, since it indicates the courts are willing to at least explore their arguments.

One of the key legal issues in the case revolves around whether the AI-generated images constitute unauthorized reproductions or derivative works of the original copyrighted images.

The artists argue that the AI models effectively contain “compressed copies” of their works, which can be reconstructed into new, infringing images. This concept challenges traditional interpretations of copyright law and could set a precedent that impacts the entire generative AI industry.

Basically, artists are saying that AI models don’t just learn from their copyrighted images, as a human might do–they allegedly embed the artist’s images into their inner workings, meaning the subsequent images they produce are just modified versions of the copyrighted originals.

Potential Impacts

If the court ultimately finds in favor of the plaintiffs, it could lead to stricter regulations and limitations on how AI companies can use copyrighted material. This could require significant changes in how AI models are trained and how companies like Stability AI and others operate, potentially “crippling” the industry if the cost of compliance becomes too high.

Companies might need to, for example, compensate artists for each image they use. Since modern AI models are trained on billions of images, this would prove extremely costly.

The case also brings up questions about the role of copyright management information (CMI) and the potential removal or alteration of CMI during the AI training process, which could violate the Digital Millennium Copyright Act (DMCA). The court seemed to find these claims less convincing and dismissed many of them.

Overall, the lawsuit underscores the fact that–while Silicon Valley might love the idea of “move fast and break things”–the vibe of the courts is more “move slowly, but be unafraid to totally upend business models.”

The case is far from decided, and the ultimate ruling could go either way. For sure, reaching an outcome will take years. But the fact that the case is even moving forward says a lot about the importance of the issues at hand.

I’ve tested thousands of ChatGPT prompts over the last year. As a full-time creator, there are a handful I come back to every day that fit with the ethical uses I mention in this article. I compiled them into a free guide, 7 Enormously Useful ChatGPT Prompts For Creators. Grab a copy today!