With the increased use of artificial intelligence (“AI”) programs, intellectual property law has been faced with a variety of unanswered questions. One such question is whether AI systems qualify as inventors for patents. Another is whether AI systems qualify as authors for copyrights. The world of trade secrets is similarly facing new questions and challenges due to the use of AI. Some of these challenges will be outlined in this post. But, first, let’s explain some basics about trade secrets.
What Makes Information a Trade Secret?
Photo by Caleb Oquendo on Pexels
First, what’s a trade secret? Some famous examples of trade secrets include the secret formula for Coca Cola, Google’s search algorithm, and the secret recipe for Kentucky Fried Chicken.
Two criteria must be met for information to be considered a trade secret. A valid trade secret must: (1) derive economic value from being classified and (2) be subject to reasonable secrecy. In addition, trade secret protection can last indefinitely. Unlike other types of intellectual property, there is no government agency for formally registering or granting trade secrets. Instead, trade secrets’ protection only exists by remaining secret. Unfortunately, it also means that if the secret gets out, it also loses its legal protection.
The reasonable secrecy requirement refers to the types of measures put in place to ensure that the information remains secret. Common measures include putting confidentiality agreements in place, limiting access to a small group of people, labeling documents as “confidential,” and using passwords to protect files containing proprietary information. So long as sufficient measures are in place, the information can still be shared in a limited way when necessary, such as to collaborate with other researchers or to merge companies.
Trade Secret Misappropriation
You may be asking yourself what legal claim exists if the trade secret is improperly revealed. Trade secret misappropriation occurs when someone acquires or discloses the secret through “improper means.” Improper means include theft, misrepresentation, bribery, or breach of a duty to maintain secrecy.
A breach of a duty of secrecy occurs when the person who discloses the information reasonably could have been expected to maintain its secrecy. A duty of secrecy is commonly imposed upon employees who have to sign non-disclosure agreements upon hiring. If an employee subsequently discloses their employer’s trade secret, they may have misappropriated it by breaching their duty of secrecy.
Owners of misappropriated trade secrets have the right to sue and seek remedies. Depending on the circumstances, an appropriate remedy may be an injunction that prevents further disclosure of the information, or it could be monetary compensation for any economic losses suffered as a result of the misappropriation.
Artificial Intelligence and Trade Secrets
The use of AI programs has greatly increased over the last few years. One widely used type of AI technology is generative AI. Generative AI refers to deep learning models that are programmed to simulate human intelligence using data and algorithms. Some popular programs that use generative AI are ChatGPT, ChatBot, Otter AI, and Grammarly.
Generative AI programs rely on users to type in information or data, known as “input,” which the AI model then uses to generate the requested content or “output.” AI programs can be trained using data from publicly available sources. There are sometimes issues about whether this training constitutes copyright infringementwhen not authorized. But, that’s not the subject of this post. Rather, an important issue is that some AI programs also use the input received from users for further training. While some people have raised concernsabout AI having access to and even storing this user input data, other users may not even be aware that their input data is being saved.
Photo by Rahul Pandit on Pexels
When trade secrets are disclosed to AI programs as input, it can cause problems for the trade secret owners. First, trade secret protection of the information could be compromised. Disclosure to AI could mean that the information is no longer subject to “reasonable secrecy” and could lose its status as a trade secret. A second problem relates to misappropriation. Even if trade secret status is maintained, disclosure of trade secrets as input to AI programs could lead to claims of misappropriation.
West Technology Group LLC v. Sundstrom – A Cautionary Tale
West Technology Group LLC recently filed a lawsuit against a former employee, Karl Sundstrom alleging trade secret misappropriation. The basic facts of this case serve as an example of the risks that AI poses to trade secrets.
Shortly after the end of Sundstrom’s employment with West, the company discovered that he’d been using Otter AI during confidential business meetings. Otter is an AI program that records and transcribes calls while simultaneously generating notes and summaries. In its complaint, West bases its misappropriation claim on the idea that Sundstrom, by recording meetings and calls, disclosed West’s trade secrets to Otter. Upon his hiring, Sundstrom signed an employment agreement which included an obligation to protect the company’s confidential information. As such, West alleges Sundstrom misappropriated West’s trade secrets when he disclosed West’s valuable confidential information in breach of his duty of secrecy.
Beyond liability for Sundstrom, West could have a separate problem. When Sundstrom disclosed trade secrets to this AI program, the information ceased being subjected to “reasonable secrecy.” While Otter claims that customer input is not stored, this may be a concern for trade secrets disclosed to AI programs that do store user input.
Can Companies Avoid this Problem?
Is there anything that companies like West can do to avoid being blindsided by employees’ unauthorized disclosures of trade secrets to AI programs? One change companies could make is to implement stricter access controls for AI. This could mean releasing specific guidelines that explain permitted versus non-permitted uses of AI programs.
Additionally, employers could have mandatory trainings to specifically educate employees about what they consider to be confidential information. These trainings could also emphasize the importance of not disclosing confidential information on AI platforms due to the risk of the software retaining the data.
For companies that want to embrace AI, it may be worthwhile to invest in developing an internal AI tool that may be trained using their own data, but not share that data with others. Then, anything submitted to the platform as input remains purely internal and will be more likely to remain confidential. As this area is developing more every day, I expect that companies will come up with creative ways to protect their trade secrets.
Callie Wershing
Assistant Blogger
Loyola University Chicago School of Law, JD 2025