Power up. From supercharging employee productivity and streamlining processes to inspiring innovation, Microsoft’s AI is designed to help you build the next big thing. No matter where you're starting, push what's possible and build your way with Azure's industry-leading AI. Check it out.
That picture of your dog as an anime character spun up by ChatGPT? And that corny diss rap you coaxed out of Claude? These are not legally protectable forms of creative expression under current copyright law.
That…could be for the best. But as businesses hand generative AI the keys to more of their marketing, content, and coding operations, what does it mean for their ability to protect the copyrights of critical business materials?
In January, the US Copyright Office released a report holding that works created with generative AI are not eligible for copyright protection without evidence of human creative contributions—and that doesn’t include an elaborate prompt.
Then in March, a federal appeals court also affirmed that art created autonomously by AI cannot be copyrighted, the latest decision in inventor Stephen Thaler’s years-long quest to copyright an output from his “Creativity Machine” AI tool.
Last month, we asked IP lawyers about where the fight around copyright protections stands when it comes to training foundation models. Now, we’re digging into the other side of the equation: questions around copyrighting AI output and what companies need to know if they’re tapping generative AI for creative and development work.
Human in the loop: Lawyers we spoke with said the safest bet for companies to ensure any AI-generated material is not replicated by a competitor is to craft policies that involve humans in creation in some identifiable way.
Sarah Bro, a partner at McDermott Will & Emery, said she advises clients using AI for content generation and marketing to consider how important it is to them that a particular piece of content not be infringed upon. If it is “material,” they need to make sure they can point to some form of human contribution.
“Is it something that is going to need to have some longevity and be reused and it’s part of this material advertising campaign for them, versus, is this a one-off run of content—you’re probably not going to use it again, and if you see something similar on the marketplace, you’re probably not overly concerned,” Bro said. “That sort of helps to draw the line, or at least help inform internal policies about how these tools can be used to generate this content, the things that your teams need to be thinking about when they’re using AI tools to generate content.”
As more companies turn to AI tools for software development work, copyright concerns may apply to that output as well. Companies should be wary of code that’s entirely AI-generated if it’s for a critical business perspective, Bro said.
“If something needs to be a more proprietary, owned, secret, then using purely AI to just crank out your new code may not be ideal,” Bro said.
Vincent Allen, a partner at Carstens, Allen & Gourley, said it’s important for now to make significant edits on any AI-generated image or text before putting it out into the marketplace.
“It means you need to make sure you keep a human touch on the finished product,” Allen said. “If you do use something like Midjourney or you use ChatGPT to write the text, then you should be making changes to it, editing so that you add enough creativity to that resulting work, that there’s no dispute that you have copyright protection in the end product.”
Will it stand? Allen recently helped craft his tech and IP law firm’s submission to the public comment period of President Trump’s forthcoming AI Action Plan. In that document, Allen and his colleagues called for more copyright protections for AI-generated works as long as there is a human creative input.
Allen said he’ll be watching the case of artist Jason Allen (no relation), who is suing to reverse a Copyright Office decision denying his AI-generated image protection. Jason Allen claims his process of hundreds of Midjourney prompts and Photoshop edits should make it eligible for protection, and that others have been stealing his work.
“That could be one of those cases that the court says, ‘I do think that there’s creative input into this resulting work. Even though GenAI created it, the author is the one that used GenAI; he controlled the output by iterating hundreds of prompts and ultimately arrived on the solution that he was happy with,’” Vincent Allen said. “That case is really not dissimilar to cases where an artist might be using drip painting, for example.”
Unlike fair use questions around training, however, Bro said she expects the Copyright Office’s reports will likely hold for the foreseeable future. “I do think that the ownership and copyrightability piece is probably going to stand for the time being,” she said.
However, Kate Downing, a lawyer focused on AI licensing and the tech industry, said she has seen a tendency in certain cases, such as the recent ruling in Thomson Reuters v. Ross Intelligence, for courts to lump together copyright questions around training with those around the output. That could create confusion that would be detrimental for businesses looking for clear rules on the topic.
“For different kinds of AIs and different kinds of AI cases, it will be very important to make one ruling related to training and another related to output, and that will be valuable guidance for how people build their companies,” Downing said. “And if you can’t distinguish the two of them, the guidance is much harder to ascertain.”