Using AI to copy photos ‘raises an interesting legal question’

Intellectual Property /
John Simpson
Latest posts by John Simpson (see all)

By Tony Poland, LegalMatters Staff • A lawsuit brought by an international visual content creator and provider against a company that uses artificial intelligence (AI) to generate computer-synthesized images could impact the future of copyright law, says Toronto intellectual property lawyer John Simpson.

In the lawsuit filed in the United States, Stability AI is accused of a “brazen infringement of Getty Images’ intellectual property on a staggering scale.”

“Stability AI has copied more than 12 million photographs from Getty Images’ collection, along with the associated captions and metadata, without permission from or compensation to Getty Images, as part of its efforts to build a competing business,” according to the court filing. “As part of its unlawful scheme, Stability AI has removed or altered Getty Images’ copyright management information, provided false copyright management information, and infringed Getty Images’ famous trademarks.”

The claim has not been proven in court. 

Simpson, principal of IP boutique Shift Law Professional Corporation, says the lawsuit “goes to the issue of intellectual property rights arising in the process of training AI.”

‘AI and intellectual property rights make for an interesting topic’

“AI and intellectual property rights make for an interesting topic. It is no longer theoretical. We are already starting to see some caselaw that charts us into new territory,” he tells LegalMattersCanada.ca. 

Getty, which markets its services to customers in more than 200 countries, claims that “rather than attempt to negotiate a license with Getty Images for the use of its content,” Stability AI scraped millions of images “even though the terms of use of Getty Images’ websites expressly prohibit unauthorized reproduction of content for commercial purposes.”

“Stability AI was well aware that the content it was scraping without permission from Getty Images’ websites was protected by copyright,” according to the claim.

Simpson, who is not involved in the case but comments generally, says the issue is likely to come down to whether the reproduction of images used to train an artificial intelligence platform is excusable under the United States law’s doctrine of fair use.

“In Canada, the concept is known as fair dealing, which basically means that anything that would otherwise be considered copyright infringement is excused on the basis that it falls under some sort of permitted use, such as research or education,” he says. “In the U.S. it is a bit different. There is a doctrine of transformative use where you can take an existing image or work and transform it into a new work. That is not something explicitly excusable in Canada but perhaps will be at some point.” 

The Getty case “raises an interesting legal question,” says Simpson.

“We are talking specifically about intellectual property rights in the training of AI and whether copyright in the materials used is being infringed in a way that it isn’t infringed in the process of training humans,” he says.

Using AI to generate images not new

The use of AI to collect information in order to generate images may not be a new concept but it raises some unexplored and interesting legal concerns, Simpson says.

“But if it is an artificial brain and there is a mechanical reproduction occurring as opposed to a biological reproduction, that is copying. The copying happens as soon as a protected image enters the AI brain. That is the issue,” Simpson adds. “Getty is saying, ‘There is inherent value in our images and we should be compensated when you use them to train your image generator.’”

“One way to look at this is to draw an analogy to how a photographer produces a good image,” he explains. “Someone might look at existing images and essentially create stock imagery in our memory that can be used to create our own artwork. Obviously, that is not copyright infringement.

AI may be able to create a new work but if it uses a copyrighted image to do it, it raises the question of infringement, he says.

“It appears the issue being framed for the court, in this case, is whether this mechanical reproduction of these training images is excusable by fair use,” Simpson says. “It would be difficult to say that it is not what would otherwise be considered an infringement. There is a reproduction happening and it is without a licence.” 

Every case is fact-specific

However, every case is fact-specific and “tiny, factual differences can make a major legal difference,” he says.

“It really matters what is going on mechanically here,” says Simpson. “There can be grey areas in the law. That is typically where something is known but it is not clear. This case is a new area more than it is a grey area. It is entirely new and much is unknown.” 

What is happening with AI in relation to intellectual property rights “is no longer a concept, it is here before us,” he says, and the issues that arise will need to be addressed.

“The answers to the many anticipated questions will come from two sources, new legislation and rulings from the highest courts in Canada and the United States,” Simpson says.

He says while artificial intelligence has been with us for some time, it is not as if lawmakers could have predicted the resulting intellectual property rights challenges.

“Copyright legislation is almost always behind the times because it is more responsive to existing developments rather than anticipatory of future developments,” Simpson says. “Should we have anticipated what is happening now? I do not believe that is fair.

“Legislation is rightly responsive to new situations,” he adds. “Where it is needed, we draft new laws and if a question comes before the courts before we have the chance to amend legislation, then it will be up to the highest courts to answer that.”