Zoe Smith
Associate Editor
Loyola University Chicago School of Law, JD 2026
Most companies own valuable trade secrets, such as the recipe for Coca-Cola or Google’s algorithm. But can a company that develops AI have trade secrets? The Uniform Trade Secrets Act defines a trade secret as “information, including a formula, pattern, compilation, program, device, method, technique, or process,” that derives economic value and is the subject of efforts to maintain its secrecy. The protection of trade secrets is essential for companies to maintain their competitive edge and drive economic growth. As such, they are instrumental in both corporate governance and compliance. Companies already deal with the risks of employees using generative artificial intelligence (AI) and exposing trade secrets; however, recent AI regulations in Europe and the United States have further increased risks relating to trade secrets.
AI regulations in Europe
Regulation (EU) 2024/1698 (The AI Act) provides “harmonized rules on AI” for the European Union. The Act requires companies keep records and technical documentation of information necessary to assess compliance of the AI system. According to the Act, such information should include: “the general characteristics, capabilities and limitations of the system, algorithms, data, training, testing and validation processes…” AI developers are pushing back on the Act, arguing that much of the information they are now required to disclose for “transparency” purposes constitutes trade secrets.
AI regulations in the United States
Unlike Europe, the US lacks comprehensive federal legislation and regulations on the use or development of AI. However, there are more than 120 bills being considered by Congress that emphasize the need for guidelines relating to AI systems. Furthermore, many states have implemented some form of regulation on AI.
While the US may not have federal legislation regarding AI, it does have legislation for Trade Secrets. Many who advocate restricting AI argue that AI may not be the creator of a work; however, the Defend Trade Secret Act does not require a human creator. Furthermore, as aforementioned, the definition of a trade secret includes all forms of information. Thus, trade secret protection could extend to the platform itself, underlying algorithms and models, training data, input parameters and model outputs. The Trade Secrets Act criminalizes the “unauthorized disclosure” of trade secrets by the federal government, but it applies only to disclosures “not authorized by law.”
Risks these regulations pose to trade secrets
A “fundamental pillar” found in AI and data protection laws is the mandate to enhance data access and sharing. In regard to the EU Data Act, that includes granting access to data to users designated by the users of the products. As of December, 2024, the EU’s AI act has provided the most significant opportunity to advance transparency around data to date. The Computer & Communications Industry Association argues that the release of the information required by the AI Act would require companies to disclose their trade secrets, stating that “publicly sharing the exact content used to train a APAI model is sensitive information” requiring strict protection.
Patent and copyright laws remain ambiguous as to their application to AI, leading many businesses to turn to trade secrets as their shield for AI. After all, nothing in the definition of a trade secret prevents the protection of information generated by AI or the system itself. Even though the information may fall under the definition of a trade secret, it must still meet the requirements to be an actual trade secret, which may be difficult for many AI systems to do. This is due to the fact that companies must take reasonable measures to keep the information secret, such as including the use non-disclosure and confidentiality agreements. Because data collection practices are constantly evolving, maintaining effective confidentiality agreements may become increasingly challenging.
Conclusion
It is important to keep in mind the major exception to trade secret disclosure – they apply only to disclosures not authorized by law. If federal regulation requires transparency of AI systems, protecting the systems as a trade secret may no longer be possible. One thing is certain, the use of AI is rapidly rising, and there is likely to be increasing regulation around it. These companies may want to (or be forced to) consider alternative ways of protecting their systems.