World's First AI Judge Assistant System Goes Live — Trial Efficiency Up 300%, But Fairness Questions Mount

World's First AI Judge Assistant System Goes Live — Trial Efficiency Up 300%, But Fairness Questions Mount

On October 1, 2027, China's Supreme People's Court officially launched the "Tianheng" AI judge assistant system nationwide, making it the first national-level system in the world to deeply embed artificial intelligence into judicial proceedings. According to data released by the Supreme Court, the system covered all 3,200 basic-level courts in its first month, processing a total of 1.87 million cases. Average case resolution time plummeted from 68 days to 17 days—a roughly 300% improvement in trial efficiency.

Tianheng was jointly developed by the Supreme Court, Tsinghua University Law School, and the Chinese Academy of Sciences' Institute of Automation over four years, with R&D investment exceeding 1.2 billion yuan. At its core is a large language model with over 200 billion parameters, trained on 45 million publicly available court rulings from all levels of the Chinese judiciary over the past 20 years, 320 laws and regulations, and more than 80,000 judicial interpretations.

How the System Works

According to Chen Mingyuan, director of the Supreme Court's Technology Center, Tianheng is not designed to replace judges but to serve as an auxiliary tool throughout the trial process. At the case intake stage, the system automatically performs case classification, evidence chain analysis, and legal statute matching. During hearings, it generates real-time trial transcripts and flags key points of contention. At the verdict stage, it provides sentencing recommendation ranges and similar-case references.

"Judges retain full discretionary authority," Chen emphasized at an October 15 press conference. "All recommendations from Tianheng are non-binding. We have designed clear human-machine interaction mechanisms, and judges can override the system's suggestions at any time."

Yet actual operations have drawn widespread concern from the legal community. In a recent paper published in Legal Studies, Professor Li Jianhua of China University of Political Science and Law reported that in a sample analysis of the first 50,000 judgments, judges adopted Tianheng's sentencing recommendations 89.3% of the time. "When a system's recommendations are accepted by nearly nine out of ten judges, can we still say judges have genuine discretionary power?" Li wrote.

Behind the Efficiency Gains

The efficiency figures are impressive. Take Beijing's Haidian District People's Court as an example. Civil division judge Zhang Wei told this publication: "A loan dispute case used to take an average of 45 days from filing to judgment. Now Tianheng auto-generates evidence analysis reports and draft judgments—I can close a case in 15 days. My daily caseload has jumped from 3 to 12."

National data from the Supreme Court show that the backlog of pending cases fell from 8.9 million at the end of 2026 to 3.1 million in November 2027—a 65% reduction. This has partially alleviated China's chronic "too many cases, too few judges" problem; the country has approximately 120,000 active judges, each handling over 200 cases per year on average.

But efficiency gains have come with new problems. A report released in November by the Human Rights Research Center at Peking University Law School found that in cases involving ethnic minorities, persons with disabilities, and migrant workers, Tianheng's recommended sentences were on average 12.7% higher than in comparable cases. "The training data itself may contain systemic biases. Social prejudices embedded in past rulings have been learned and amplified by the algorithm," said the report's lead author, Professor Wang Min.

International Attention

The system has also drawn close scrutiny from the international legal community. On October 20, the UN Office of the High Commissioner for Human Rights issued a statement expressing concern over the "transparency and accountability" of China's AI judicial system. "Any artificial intelligence system that assists judicial decision-making must meet the basic requirements of due process, including algorithmic explainability and the right of parties to be informed," the statement read.

Ragnar Hedlund, chair of the European Commission for the Efficiency of Justice (CEPEJ), said at an online seminar: "China's experiment offers us an observational window. The balance between efficiency and fairness is a core issue for all judicial systems, and AI involvement makes it more complex."

In response, the Supreme Court issued on November 5 the Provisions on Transparency in the Operation of the Tianheng System, requiring courts at all levels to inform parties of AI assistance usage and granting parties the right to request manual review. But Sun Xianzhong, a researcher at the Chinese Academy of Social Sciences' Institute of Law, believes these measures remain insufficient. "Parties are already in an information asymmetry when facing a judge. Asking them to proactively request review is difficult to implement in practice," Sun told this publication.

The Debate Continues

The controversy surrounding Tianheng is far from settled. Supporters argue that AI assistance can effectively reduce inconsistent sentencing across similar cases and enhance public trust in the judiciary. Opponents worry that over-reliance on algorithms will erode judges' professional competence and legitimize systemic bias.

Professor Liu Xianquan of East China University of Political Science and Law raised a deeper question: "When the public knows that AI plays a role behind a verdict, will their trust in the judiciary increase or decrease? We currently lack the social psychology research to answer that."

The Supreme Court plans to complete its first comprehensive evaluation of the Tianheng system by March 2028. Until then, the debate over technology and justice will continue.