The Disney-OpenAI deal and generative AI copyright concerns

A partner at the law firm Davis Wright Tremaine LLP, Driscoll joined the Digiday Podcast to draw some of the legal lines surrounding brands and generative AI (Editor’s note: This episode does not constitute legal advice). To be clear, those lines are far from clean and fuzzy enough as to be almost imperceptible. At the same time, though, what’s become clear is generative AI is becoming a more widely adopted technology among marketers.

“Early on there was a lot of activity around brands instructing their ad agencies, their design firms, production companies, ‘Do not use any generative AI tools in connection with our work.’ That has kind of flipped, such that now a lot of companies are embracing it. And instead of prohibiting the use of generative AI, they’re mandating it,” said Driscoll.

That’s not to say every brand is outsourcing their entire marketing creative work to ChatGPT or Midjourney. Disney may have gotten comfortable enough to secure a deal with OpenAI to allow for people to create AI-generated videos of Mickey Mouse and Darth Vader. But the company likely had a platoon of lawyers looking at how such a deal could risk diluting its trademarks.

“One of the trademark law principles is that trademarks have to be used consistently and in a manner that’s under the control of the trademark owner, essentially,” said Driscoll. “So if a trademark owner takes their logo, for example, and allows it to be adapted, modified and animated by other people who aren’t under the control of the brand, there at least is a technical legal argument that that’s actually diminishing the legal protection that’s available for that trademark.”

Here are a few highlights from the conversation, which have been edited for length and clarity.

The unknown of AI-generated copyright infringement

In the United States, the way copyright law works, if a person or company generates something new that is basically the same as an existing piece of work, that could be copyright infringement, even if there’s not an intentional copying happening. And so that’s really the fear with generative AI: that a company is going to use an AI tool to produce some creative material and just not know that, in fact, the output is closely similar to something that already exists out there. That is still a concern.

A method for managing copyright concerns

Most companies that are using AI think it’s important that the generative AI output is more of a starting point as opposed to the end product. So there’ll be some human intervention, human creativity applied. It’s not foolproof obviously. But the idea that the generative AI output might be raw material but not the final product, and the final product is going to reflect some other significant creative input from a person is part of how most companies that I’m aware of use generative AI.

A range of risk tolerances

There may be some brands who just say, “Look, we don’t want to produce advertising material that we can’t own.” Whereas others might say, “We get it. We may not own this material. In fact, the AI tool may spit out similar material for other users, but we’re okay with that because we like the output or we like the message, and we just understand that’s part of the picture that you get when you use generative AI.”

The pro of private licenses

Many companies – at least the ones that are most sophisticated that are using generative AI – they’re not using off-the-shelf, publicly accessible tools. They’re using enterprise software that they have licensed. And usually one of the terms of a license is that their inputs are not going to be used to train a model or to generate output for other people, for other users of the platform.