AI-generated content has unlocked affordable opportunities for small-business owners to quickly expand their reach. However, the question remains: Do you actually own the content produced by AI? The answer isn’t straightforward.
Content that’s entirely generated by AI, without any human editing, is not eligible for copyright protection and is considered public domain. This means that if your insightful blog post was written by ChatGPT, it can be republished word for word by a competitor. The real question is, how much human input is needed to claim ownership of AI-generated content? The legal framework around intellectual property (IP) in AI is still developing and is continually evolving.
To help you navigate this complex landscape, this post will cover the basics of intellectual property and copyright as they relate to AI, and it will offer insights into responsibly integrating generative AI in your business. We’ll dive into a case study featuring a fictional gluten-free bakery’s branding elements created with AI and uncover the challenges it faced. Finally, we’ll examine some pending lawsuits against OpenAI, which are expected to help clarify and shed light on the evolving legal framework for AI content.
This blog is based on a live Masterclass by IP attorney Elizabeth Milian, a managing partner at The Milian Legal Group, where she focuses on brand protection strategies and trademark law.
There are two main IP considerations as a small business. First, you want to make sure that your key intellectual properties (brand, logo, trade secrets) are protected through trademarks and copyright. Second, you don’t want to incorrectly claim as your own IP any AI-generated content you use to promote your business. Below are the key takeaways from Elizabeth’s session.
Branding, including elements like a business’s name and visual assets, such as logos, holds significant value for any company. These brand elements play a crucial role in achieving sales and driving revenue. Strong branding also helps businesses stand out in both the online marketplace and brick-and-mortar settings. While the law provides special protections to safeguard your brand, it’s essential to take the correct steps to ensure those protections are in place.
Be cautious when using free online naming tools, because they often generate “brandable” names or logos based on basic prompts to create “unique” or “creative” names for your business. Elizabeth explored this by testing a hypothetical business – a gluten-free bakery – on five different free naming sites.
When using AI for brand creation, it’s essential to consider the potential intellectual property issues tied to AI-generated outputs. Large language models like ChatGPT are trained on extensive datasets, often including scraped content that may contain IP-protected works. Several copyright infringement cases are pending that highlight these issues. Free AI naming and logo-generation tools carry certain risks, so it’s important not to use their outputs without doing the due diligence to ensure originality and avoid IP conflicts. To conduct due diligence on a potential business name, start by searching the USPTO website to see if the name is already in use or registered by another business. If you find a name that’s similar, it’s wise to consult an IP attorney for a professional evaluation to ensure there’s no potential for infringement or conflict. An attorney can provide guidance on how closely your name aligns with existing trademarks and help you navigate any necessary adjustments.
Holding a copyright grants the author or copyright holder an exclusive set of rights over their work. These rights include the ability to reproduce the work in copies, create derivative works (adaptations based on the original), and control public distribution through sale, transfer, rental, lease or lending. Additionally, copyright provides the right to publicly perform and display the work in various settings.
Are works generated entirely by AI, with no human authorship, eligible for copyright protection? The answer is no – human authorship is a fundamental requirement for any copyright claim. (This was affirmed in Thaler v. Perlmutter (Register of Copyright), U.S. District Court DC, Aug. 18, 2023.)
Content created solely by AI is not eligible for copyright protection, meaning AI-generated works hold no exclusive rights and are considered part of the public domain – free for anyone to use, transform or display.
Here’s an example. If a business publishes blog posts using AI-generated outlines as a starting point, is the content protected? The answer depends on the originality contributed by the author. If the author adds creative wording, and selects and arranges information in a unique way, those elements may meet the originality requirement and be protected. However, only the new, original contributions are eligible for protection; there is no exclusive right over the preexisting material generated by AI.
Alter v. OpenAI Inc., 1:23-cv-10211, (S.D.N.Y.): In several ongoing cases, more than a dozen prominent writers, including John Grisham, Jonathan Franzen and Jonathan Alter, are alleging that OpenAI and Microsoft are liable for copyright infringement. The core argument centers around unauthorized use: OpenAI allegedly used copyrighted works without permission to train its GPT models. Plaintiffs claim this use has caused economic harm, arguing that the unauthorized use of their works has resulted in financial losses. They contend that, without access to these copyrighted materials for training, OpenAI and Microsoft would not have a viable commercial product. This has, they argue, unfairly damaged the market for professional authors’ works.
The New York Times v. OpenAI 1:23-cv-11195, (S.D.N.Y.): The New York Times claims that millions of its copyrighted works were used to train Microsoft’s Copilot (formerly Bing Chat) and OpenAI’s ChatGPT without authorization. According to the NYT, its copyrighted works were both directly and indirectly infringed by being included in training datasets. The AI tools allegedly generate outputs that recite the NYT’s content verbatim, closely summarize the content, mimic the expressive style of the NYT and sometimes even falsely attribute outputs to the NYT.