And here is the next (very timely!) post in our Chinese arbitration series:

AI Inside Arbitration Institutions

Guan Yunbiao

Much of the current discussion about artificial intelligence and dispute resolution focuses on individual users: whether lawyers should use generative AI to draft briefs, whether arbitrators may use it for legal research or award drafting, and what disclosure should be required. Developments in China point to a different question: what happens when AI is incorporated into the arbitration institution itself?

Several leading Chinese arbitration centers already are incorporating AI into the infrastructure through which cases are filed, administered, heard, and decided. Their experience offers an early look at how institutional AI may change arbitration and at choices the broader ADR community may soon face.

From Online to Intelligent Arbitration

China’s adoption of AI builds on a broader digitalization of arbitration. Major Chinese centers began using online filing, electronic service, and virtual hearings before the COVID-19 pandemic, which accelerated those developments. What began as an emergency response has become permanent institutional infrastructure. The revised Arbitration Law, effective March 1, 2026, expressly recognizes online arbitration and gives online proceedings the same legal effect as offline proceedings, subject to party objection.

Once an institution has digitized a case from filing through award, AI becomes a natural next step. The platform already contains pleadings, evidence, hearing records, and procedural communications. AI can organize, summarize, translate, transcribe, retrieve, and analyze that information. Several Chinese institutions are testing what this can mean in practice.

The Shenzhen Court of International Arbitration was an early mover. In 2016 it introduced its “3i Robot,” an AI-based system designed to answer arbitration-related questions for lawyers and arbitrators. Shenzhen later disclosed research using large collections of arbitral awards to develop tools that could assist arbitrators in drafting and reviewing awards.

The Guangzhou Arbitration Commission has integrated AI more deeply into case administration. Its “Yun Xiaozhong” system provides arbitration information and supports 24-hour intelligent case filing. “Zhong Xiaowen” functions as an AI arbitration secretary, with reported capabilities including multilingual translation, evidence verification, automated transcription, recording party statements, and generating draft awards from hearing records. In February 2024, Guangzhou’s L-Code full-process AI-assisted arbitration system assisted in mediating a loan-agreement dispute before the hearing.

In April 2025, the Shanghai International Arbitration Center launched a large-language-model-based AI-assisted arbitration system. It can extract claims, evidence, and cross-examination materials from case files, organize them into standardized formats, and identify important factual and legal differences between the parties. The objective is not to have AI decide disputes, but to reduce the time arbitrators spend organizing information so that human decision-makers can concentrate on adjudication and award writing.

These systems differ in sophistication and function, but they illustrate the same shift: AI is moving from a tool an individual participant might choose to use into part of the institutional machinery through which arbitration is administered.

Assistance, Not Adjudication

That shift also helps explain an important boundary emerging in China. Chinese policymakers and arbitration institutions generally have not embraced autonomous AI adjudication. The Supreme People’s Court has stated in the judicial context that AI should assist rather than replace judges. In July 2025, the China International Economic and Trade Arbitration Commission (CIETAC) issued AI guidelines built around party autonomy, good faith, and an “assistive adjudication” principle. The Institute of International Law of the Chinese Academy of Social Sciences also has issued recommended guidelines on AI use in arbitration.

The immediate applications therefore are not machines independently deciding who wins. They include filing, translation, document organization, transcription, legal-information retrieval, issue identification, and preparation of draft materials for human review. This suggests an alternative to framing the issue as a choice between rejecting AI and permitting an “AI arbitrator.” Technology can perform administrative and analytical functions while responsibility for adjudication remains with human arbitrators.

Institutional AI Raises Different Questions

Embedding AI within an arbitration institution creates issues different from those arising when an individual lawyer uses a public generative-AI tool. Transparency is one. Parties know that an institution will assign case managers and perform administrative functions, but should they be told which portions of that work are AI-assisted? Does the answer change if the technology schedules a hearing rather than summarizes evidence? CIETAC’s guidelines provide one possible approach by permitting tribunals, where appropriate, to invite party views on AI use.

Reliability is another concern. A mistaken automated reminder is inconvenient; an inaccurate AI-generated summary of testimony or evidence is much more serious. The closer AI moves toward substantive case analysis, the more important human verification, auditability, and clear allocation of responsibility become. Chinese practice to date retains human responsibility for the resulting decisions and actions.

Confidentiality and data security present another challenge. Institutional AI may have access to pleadings, contracts, trade secrets, witness testimony, personal information, and draft awards. Institutions therefore must address where case data are stored, how the system is trained, who can access the information, and whether case materials can be used beyond the particular proceeding.

Technological capacity also is uneven. China’s largest arbitration institutions can develop proprietary platforms and experiment with sophisticated AI systems; many smaller institutions cannot. If AI reduces cost or improves case administration, unequal access to technology could become a new source of institutional competitive advantage.

A More Useful Question

The Chinese experience suggests that the familiar question—whether AI will replace arbitrators—may be less useful than asking which components of the arbitral process institutions should delegate to or perform with AI, and under what safeguards. Filing and scheduling are relatively easy cases. Translation, transcription, evidence organization, and issue identification move closer to adjudication. Drafting an award moves closer still.

Chinese arbitration centers are experimenting along this continuum. Other institutions will make different choices, shaped by their legal rules, professional norms, technological capacities, and user expectations. The Chinese experiments allow the international ADR community to watch those choices being made in real time.

The future of AI in arbitration may therefore arrive not when an AI system takes the arbitrator’s chair, but much earlier—when parties discover that AI already has become part of the institution sitting behind it.

About the author

Guan Yunbiao is a PhD candidate at Peking University Law School.

This post draws on Guan Yunbiao, “Arbitration in the Digital Age: Technological Innovations in Chinese Centers,” in Richard Bales & Dan Xie, eds., The Cambridge Handbook of Arbitration in China (Cambridge University Press, forthcoming spring 2027). The submitted chapter is available on SSRN.