Despite legal and ethical concerns, many business leaders are still bullish on generative artificial intelligence (AI), and the industry is accelerating at a rapid pace. In fact, according to a UBS study, ChatGPT reached over 100 million monthly active users in the two months following its initial launch—making it the fastest-growing consumer application in history.
Unfortunately, as the use of generative AI surges, so too have the legal questions surrounding it. Copyright and trademark disputes regarding AI-generated material are on the rise—thrusting intellectual property (IP) law into uncharted territory. The cases being decided today, barely two years into the generative AI boom, may establish legal precedent that shapes the future of law for decades to come.
In the past several months, several high-profile legal cases have tested the boundaries of IP in the age of artificial intelligence, highlighting key issues.
Copyright Infringement When Using Content Created by AI
In Alcon Entertainment, LLC v. Tesla, Inc., Alcon Entertainment, the exclusive rights holder of the 2017 film Blade Runner 2049, has accused Tesla and Warner Bros. Discovery of using AI-generated imagery that closely mimics an iconic image from its film without prior permission. The image—depicting a man next to a futuristic-looking vehicle—was used at Elon Musk’s Cybercab launch event.
The suit claims that Tesla initially requested to use an actual image from the film. When Alcon Entertainment denied the request, the suit alleges Tesla and Musk used generative AI to create similar visuals, using Blade Runner images as a close reference.
Interestingly enough, whether a work is generated by AI or created by humans is irrelevant in this case. The question of fact will remain the same irrespective of how the image was created: Does this work infringe on Alcon Entertainment’s intellectual property?
But this case still provides attorneys with a valuable lesson. We will likely see these types of cases grow exponentially. AI tools can now produce images or written works in seconds, much more quickly than a human could. However, the images are not created from the “mind” of the AI; they are based on the information the AI has been exposed to already. Therefore, the “new” content the AI is creating is, most of the time, based on a human’s work.
Many day-to-day users of generative AI platforms are unaware of the potential risks associated with how these tools are trained. They may believe that the work they have asked generative AI to create is truly “original.” In many instances, individuals may not realize that their generated content mirrors copyrighted material until it’s too late.
For this reason, IP attorneys must continue to educate their clients about the risks involved in using generative AI and encourage them to rely on original, human-created content when possible. While AI can certainly aid the creative process, it should not be relied on as the primary source for public-facing materials. Of course, any AI-generated works should be reviewed by legal experts with scrutiny, as any prudent attorney would with work created by a human.
Copyright Infringement When Using Content to Train AI
The rapid adoption of generative AI has given rise to legal disputes not only about the output of these tools, but also concerning how these tools are trained.
For example, in Toronto Star Newspapers Ltd. v. OpenAI, Inc., more than five of Canada’s most prominent news outlets, including the Toronto Star newspaper, have filed a suit against OpenAI, alleging that OpenAI “scraped” content from their websites to train ChatGPT without their consent. The plaintiffs argue that this constitutes copyright infringement.
Similarly, in another landmark case filed this June, UMG Recordings, Inc. v. Uncharted Labs, Inc., several major record labels have teamed up to file a suit against Uncharted Labs for copyright infringement. The plaintiffs allege that the company’s AI model, Udio, illegally copies digital sound recordings to train its system. The tech then generates music that imitates the qualities of genuine, human-made recordings.
One potential outcome of both of these cases could be the establishment of licensing agreements. In the early days of music streaming platforms, similar legal battles ensued. The result? Licensing agreements that compensated artists for their music. Such license agreements laid the foundation for today’s major streaming platforms like Spotify, Apple Music, and Pandora. Today’s legal battles could similarly shape how generative AI companies acquire legal consent to use content generated by human artists in their training models.
For example, newspapers could provide AI companies with a license to scrape their news articles so long as the company pays a fee for the content and links back to the newspapers’ original article to credit the source of the answer or new material created. OpenAI has already announced licensing deals with The Associated Press and News Corp, among others.
Likewise, owners of music could potentially provide an AI platform with a license to train on their songs and create new music so long as a royalty is paid for using the music and users are notified that using any “new” music is subject to the copyright rights of the original owners.
When it comes to generative AI, Pandora’s box has been opened. Like it or not, this technology is here to stay. For this reason, attorneys should continue to closely monitor the ongoing cases surrounding generative AI to keep abreast of the rapidly evolving legal landscape. As the legal framework solidifies, those who stay informed will be best positioned to serve their clients.
