Dec 1, 2023 - Technology

Copyright law will shape how we use generative AI

Illustration of a confused-looking robot with copyright symbols for eyes

Illustration: Sarah Grillo/Axios

In the year since the release of ChatGPT, generative AI has been moving fast and breaking things — and copyright law is only beginning to catch up.

Why it matters: From Section 230 to the Digital Millennium Copyright Act (DMCA) to domain name squatting protections, intellectual property law has shaped the internet for three decades. Now, it will shape the way we use generative AI.

Driving the news: The Biden administration's recent executive order contained no initial guidance on copyright law and AI, which means these decisions will largely be left up to the courts.

Between the lines: In August 2023, a federal court in Washington, D.C. ruled that copyright law only extends to humans, meaning that any work of art created with AI and without human input can't be copyrighted.

  • Once humans alter that AI-created material in any way, the legal waters get murky — and we still have no idea how the law will handle disputes over attribution and disclosure.
  • "Copyright is incredibly broad and all-encompassing," technology lawyer and podcaster Denise Howell tells Axios. "We're in the wild west days of this, as far as how it's going to shake out legally."

Flashback: When search engines first launched, litigants sued them, too.

  • "Search does a lot of copying. And it's necessary to copy and index web pages in order to be able to search them," says Howell.

Erica Han, intellectual property partner at Ropes & Gray, disputes the claim that existing copyright rules don't answer novel questions about originality and authorship posed by AI.

  • "U.S. copyright law has proven quite versatile to adapting to new technologies," Han told Axios. "Similar sentiments arose throughout history after the creation of technologies such as cameras and VCRs."
  • "Yes, generative AI poses new and very interesting copyright questions," Han said. "The basic principles of copyright law are sound."

What they're saying: "I think with respect to copyright law, you will see people with legitimate claims financing lawsuits and litigation," former President Barack Obama told The Verge.

  • "Through the courts and various other regulatory mechanisms, the people who are creating content are going to figure out ways to get paid and to protect the stuff they create. It may impede the development of large language models for a while, but over the long term, that'll just be a speed bump," Obama said.

The latest: Last month, a group of nonfiction writers brought a class action complaint against OpenAI and Microsoft, accusing the companies of using copyrighted materials to train ChatGPT without authors' permission.

The big picture: Last month a federal court dismissed the copyright infringement claims against Midjourney and DeviantArt, calling them "defective in numerous respects."

  • The class action lawsuit deals with copyright issues over the art used to train generative AI tools (input) and the art they produce (output).

The intrigue: The judge who ruled in that case, William Orrick, also ruled in 2016's "monkey selfie" case that copyright can be granted only to humans, and not to animals.

  • That decision could serve as a precedent in determining how the law should deal with copyright claims for AI-generated works.

Yes, but: Enforcing an artist's copyright doesn't guarantee the artist will profit from their work.

  • The music industry could win its recent lawsuit against Anthropic for using copyrighted lyrics, but the division of any revenue from a settlement is hard to predict.

Author and activist Cory Doctorow argues that copyright law isn't the answer to the generative AI problem.

  • "We've spent 40 years expanding copyright. We've made it last longer; expanded it to cover more works, hiked the statutory damages for infringements and made it easier to prove violations," Doctorow writes.
  • "This has made the entertainment industry larger and more profitable," Doctorow argues, "but the share of those profits going to creators has declined, both in real terms and proportionately."

The U.S. Copyright Office posted a notice of inquiry and recently published submitted comments on AI-related questions, including the use of copyrighted works to train AI models, the ability to copyright anything generated with AI, and how to treat AI that's designed to imitate the the style of human artists.

  • A VP at Stability AI quit his job, declaring on X, "I don't agree with the company's opinion that training generative AI models on copyrighted works is 'fair use'."

One battle to watch is over people's likenesses.

  • Celebrity impersonators are nothing new, but generative AI now makes it easy to impersonate with realistic detail on a massive scale.
  • According to Variety, Scarlett Johansson has "taken legal action" against an AI company that used a deepfake version of her to advertise its product.
  • Likeness laws vary from state to state, but a new bipartisan bill called the No Fakes Act aims to prevent companies from using celebrities' likenesses without their consent.

What's next: We're about to enter a golden age for copyright trolls, legal experts say.

  • Just as patent trolls extract money from companies who create commonly used technology, copyright trolls will buy rights from creators or create works "solely for the purpose of going out and enforcing the copyright on them and shaking people down," says Howell.
  • "There will be software ecosystems polluted with proprietary code that will be the subject of cease-and-desist claims by enterprising firms," Sean O'Brien, founder of the Yale Privacy Lab, told ZDNet.
Go deeper