AI Synthetic Witness System SynthWitness Piloted in Courts: Can AI-Generated Testimony Serve as Evidence

Article

The Shanghai Intellectual Property Court announced in November 2028 a pilot program introducing SynthWitness — an AI-based synthetic witness system. The system integrates surveillance video, communication records, IoT sensor data, and other sources, using AI reasoning to generate event reconstruction reports as auxiliary evidence for judicial proceedings.

SynthWitness functions like a superhuman analyst. After receiving case-related data, the system reconstructs event timelines, causal relationships, and key details through multimodal fusion reasoning. Generated reports are presented in structured format, including event summaries, evidence chain analysis, and uncertainty assessments.

In the intellectual property case pilot, SynthWitness generated event reconstruction reports across 200 cases, with 92% adopted by judges as valid auxiliary evidence. Average case processing time was reduced by 40%.

"SynthWitness doesn't replace witnesses — it compensates for their limitations," explained Vice President Shen Zhixian of the Shanghai IP Court. "Human witnesses forget, are influenced by emotion, and can lie. SynthWitness is based on data and isn't affected by these factors."

However, the legal community has raised persistent questions. Professor Ma Changshan of East China University of Political Science and Law noted: "AI-generated reports appear objective, but the algorithm itself may contain bias. If training data is skewed, the resulting reports will also be skewed — but this bias is harder to identify and correct than human witness bias."

Another point of contention is liability. If SynthWitness's report leads to an incorrect judgment, who bears responsibility? The system developer, the court using it, or the training data provider? China's Supreme Court has established a special working group to study the legal framework for AI evidence.

Internationally, U.S. federal courts maintain a cautious stance toward AI-generated evidence, requiring that AI system reasoning processes be explainable. The EU is more aggressive — a draft AI Evidence Law under development would require all AI evidence to undergo independent third-party verification.