Professor Hrdy has posted a new article, forthcoming in the Stanford Journal of Law, Economics, & Business, and co-authored with Joseph Avery (Miami) and Mike Schuster (Georgia), which strongly suggests that there is an “anti-AI bias” in trade secret law.
A draft of the article, entitled The AI Penalty in Trade Secret Law, can be downloaded on SSRN.
In trade secret cases, judges and jurors may be more likely to find that a defendant “misappropriated” trade secrets if artificial intelligence (AI) was used to accomplish the task. In a leading case, Compulife v. Newman, the Eleventh Circuit found that using automation (nonhuman means) to access trade secrets was more likely to be “improper” under trade secret law than using humans to do the same thing. In the article, Prof. Hrdy and her co-authors use surveys to determine whether jurors and other factfinders are more likely to find that access was “improper” when defendants used AI to obtain the trade secrets. They showed survey respondents a hypothetical fact pattern (the “AI condition”) in which a competitor uses AI to extract trade secrets from the plaintiff’s website. They also gave other respondents a “control” fact pattern where the competitor used human means (a team of workers) to do the same thing manually.
The results? Participants in the AI condition were significantly more likely to find that the competitor had misappropriated the plaintiff’s trade secrets. When asked to issue a binary verdict on liability, 73.3% of participants in the AI condition concluded that the defendant should be held liable, compared to 60.3% in the human condition—a statistically significant difference.
Profs. Hrdy, Avery, and Schuster argue that trade secret law is much more conducive to anti-AI bias than other areas of IP law, especially patent law, because trade secret law often requires fact finders to assess whether the defendant used a “bad act” to access the secrets. Courts tell jurors to assess, in determining which acts are “improper,” both the “commercial morality” of the defendant’s actions in the industry and context, and the degree to which the plaintiff could have foreseen and prevented the defendant’s actions. Utilizing automation and novel artificial intelligence tools is currently viewed as an advanced technique for accessing others’ information; whereas trade secret holders might be expected to prevent manual means of access, they cannot necessarily be expected to prevent the use of AI.
A draft of the article, entitled The AI Penalty in Trade Secret Law, can be downloaded on SSRN.