Trade Secret Law and AI: Navigating emerging risks and legal implications

The age of artificial intelligence (AI) is here, and the impact of generative AI on different areas of law is the subject of much debate. For trade secret law in particular, AI could be a double-edged sword: it is a tool that could enhance trade secret protection by improving monitoring, employee training, and other precautions against improper disclosures of proprietary information, but could also threaten trade secrets by potentially raising the standard for protection and reducing the effectiveness of certain remedies in cases of misappropriation. While it remains to be seen how courts address these issues, we explore some potential new considerations AI may pose for trade secret protection and remedies.

AI AS A POTENTIAL THREAT TO TRADE SECRET PROTECTION

Under the Uniform Trade Secrets Act (UTSA), which has been adopted in nearly all states, a trade secret is "information" that has independent economic value as a result of it "not being generally known" and "not being readily ascertainable by proper means" by others who can "obtain economic value from its disclosure or use." UTSA Section 1(4). A trade secret owner must make "efforts that are reasonable under the circumstances" to maintain its secrecy, or it will no longer qualify for protection. See id.

The Defend Trade Secrets Act (DTSA) sets the criteria for trade secret protection under federal law and echoes the state-based requirements. 18 U.S.C. Section 1839(3).

AI could be used by litigants to try to undermine trade secret protection in a number of ways.

"READILY ASCERTAINABLE" INFORMATION

Defendants may argue that generative AI's ability to ingest complicated inputs and produce sophisticated outputs makes it more difficult to protect certain information as a trade secret because such information has become more "readily ascertainable." For example, there is a risk that AI can be used to quickly analyze vast troves of data to reverse engineer confidential information such as customer lists, chemical formulas, or proprietary methodologies. Judges and juries may also come to view the use of AI as "proper means" for ascertaining or acquiring trade secrets, especially as AI becomes more widely used. Or AI could simply increase the availability of new and valuable information such that fewer things qualify as "secret" (i.e., not generally known and not readily ascertainable) in the first place.

"REASONABLE EFFORTS" TO MAINTAIN SECRECY

Defendants may also argue that the standard for "reasonable efforts" to maintain secrecy requires more extensive precautions in the age of AI given the risk that a trade secret owner could lose control over information once it is exposed to certain AI models. Trade secret owners may need to mitigate this risk by using advanced security measures or "proprietary" AI tools or modes that segregate confidential information from model training or public use. Another open question is to what extent, if at all, trade secret owners need to include explicit restrictions on AI use in their confidentiality agreements to demonstrate their "reasonable efforts" to keep information secret.

AI'S POTENTIAL IMPACT ON TRADE SECRET REMEDIES

AI also poses potential implications for trade secret remedies, including unjust enrichment and apportionment of damages.

UNJUST ENRICHMENT

Both the UTSA and the DTSA allow for recovery of unjust enrichment damages that are not addressed in the computation of damages for actual loss. UTSA Section 3(a); 18 U.S.C. Section 1836(b)(3). There have been several large awards recently, mostly based on theories of unjust enrichment such as recovery of a defendant's sales, avoided development costs, and associated head-start damages. See, e.g., Appian Corp. v. Pegasystems Inc., No. 2020-07216 (Va. Cir. Ct. 2022) ($2.04 billion) (appeal pending); Syntel v. TriZetto, 68 F.4th 792, 814 (2d Cir. 2023) (vacating $285 million avoided costs award as "unavailable under the specific facts of the case"); Epic Sys. Corp. v. Tata Consultancy Servs., Ltd., 980 F.3d 1117, 1132-33 (7th Cir. 2020) (upholding $140 million award based on avoided costs and "head start").

As AI adoption accelerates, there is a risk that it will enable competitors to exploit trade secrets more rapidly, which could impact the scope of unjust enrichment claims. In some cases, plaintiffs rely on the time, money, and effort they spent developing a trade secret as a proxy for the time, money, and effort that defendants avoided by misappropriating the trade secret. However, defendants may argue that the speed or effectiveness with which AI facilitates the development of new and valuable information lowers avoided costs and shortens the effective head-start period associated with certain trade secrets.

APPORTIONMENT OF DAMAGES

Some courts have required that damages be apportioned to the amount attributable to the relevant trade secrets. The Sedona Conference, Commentary on Monetary Remedies in Trade Secret Litigation, 24 SEDONA CONF. J. 349, 407-408 (2023). This may be the case where a plaintiff seeks damages for lost profits, lost sales, or disgorgement of the defendant's profits, as some of the sales or profits (e.g., in a multi-component product) may be unrelated to the alleged misappropriation or trade secrets at issue.

AI-driven software and processes may help trade secret owners trace development time and expenditures related to specific trade secrets. But as AI becomes more commonplace in the research and development process, it may also become more difficult to parse how much of a defendant's sales or profits are due to individual trade secrets versus AI's more general capabilities, insofar as AI models are "black boxes" that do not provide clear insight into how or why they reached certain results or generated certain output.
• • • • •
AI's intersection with trade secret law is an area of uncertainty that will continue to evolve as technology advances. While AI does not necessarily change the standards for trade secret protection and remedies, both plaintiffs and defendants should be alert to the possibility that developments in AI, including generative AI, could have an impact on trade secret law, including protectability and remedies.

Miriam Kim and Vincent Y. Ling are partners, and Nathaniel F. Sussman is an associate at Munger, Tolles & Olson LLP. Kim and Sussman also co-chair the firm's Generative AI Task Force.