ABSTRACT
Factfinders are significantly more likely to assign liability, award higher damages, and judge conduct as improper when defendants use artificial intelligence (AI) systems-rather than humans-to acquire trade secrets, even when the conduct is otherwise identical. This finding, based on original experimental research, reveals that there is an ‘AI penalty’ in trade secret misappropriation cases. In other IP regimes, anti-AI bias is doctrinally inappropriate: the method of infringement should not affect adjudicatory outcomes. But trade secret law is different. Its open-textured standards-such as ‘improper means’, ‘reasonable measures’, and ‘commercial morality’ – demand fact-specific, context-sensitive analyses. As a result, courts may permissibly, even appropriately, reach different conclusions about AI – and human-enabled conduct. Still, when courts penalize AI simply for being AI, they risk codifying a de facto rule that AI and automation are inherently suspect. This is problematic. Normatively, it may chill legitimate uses of AI or incentivize firms to conceal their methods of information gathering. Doctrinally, it undermines the principles behind trade secret law’s evolving standards, which must track shifts in technological capabilities and industry adoption. Thus, we outline how courts can articulate clear, principled reasons when treating AI-enabled conduct as improper – preserving the adaptability of trade secret law while ensuring that its evolution remains responsive to, rather than distorted by, new technologies.
Schuster, W Michael and Avery, Joseph and Hrdy, Camilla Alexandra, The AI Penalty in Trade Secret Law (June 1, 2025), Stanford Journal of Law, Economics & Business (forthcoming 2026).
Leave a Reply