AI Litigation Focus Shifts from System Construction to Behavior

In Q3 2026, J.S. Held's AI Disputes Monitor highlights a shift in AI litigation focus. Forty-two new cases broaden the scrutiny from training data to AI system behaviors, examining outputs, product safeguards, and corporate oversight. The dataset now includes 468 cases, with 115 filed in 2026. Early litigation focused on the materials for training models, while current cases look at AI conduct post-deployment, assessing its operational consequences.

AI-related legal disputes now examine broader aspects, including model developers, platforms, and companies using AI, expanding across the AI value chain. The evidence needed for claims evaluation has changed, requiring a look at system logs, integration records, and governance documentation. The Q3 analysis highlights that technical causation and economic harm often rely on the same factual evidence, emphasizing the need for coordination among technical and governance experts.

Two forms of AI disclosure risk have emerged: overstatement of AI features and underestimation of AI's impact on business operations. Directors, investors, and counsel face risks both from overstating AI capabilities and underestimating competitive effects. The Q3 additions stress AI litigation’s need for comprehensive and coordinated expert analysis to address technical, economic, and governance questions.