When the System Is the Witness

Lou Schwartz, Esq., Managing Member, Schwartz & Associates

Litigation has always relied on human accounts. What was known. What was decided. Why a choice was made. Artificial intelligence changes that framework.

Modern AI systems generate contemporaneous records of decision-making. Inputs are logged. Outputs are timestamped. Models apply rules consistently. When disputes arise, courts are no longer limited to reconstructing intent through testimony. They are presented with system behavior.

This raises a threshold evidentiary question. When outcomes repeat across a population, does the system’s output itself establish conduct.

Courts are already moving in this direction. Algorithmic decision tools used in pricing, screening, and content moderation are increasingly treated as centralized decision-makers rather than passive instruments. Where a system applies uniform logic, plaintiffs argue that its outputs supply common proof.

This does not eliminate the need for human accountability. It relocates it. Design choices, training data, deployment parameters, and monitoring practices become the focus. The relevant question is no longer what an individual actor intended, but what the system was built to do.

As AI-generated evidence becomes more common, courts will need to decide how to weigh consistency against context. Whether repeated outcomes establish liability. Whether opacity cuts for or against admissibility. And how much explanation is required before a system’s output can be credited.

Those answers will define the next phase of complex litigation.