MMOOZZGG.com

AI for lawyers

AI in the justice system: how countries let AI into the courts

6 min read9 sources

AI in the justice system already works in China, England, Colombia and the EU, and since September 2026 Russia has its own Supreme Court concept too. The approaches differ in timing and strictness: China is building rules for the whole process by 2030, and the EU puts AI for courts in the high-risk class, but everywhere the decision stays with the judge. Below is a comparison by country and what follows for lawyers.

On this page · 5 sections

China

In December 2022 the Supreme People’s Court of China issued its “Opinions on Regulating and Strengthening the Applications of AI in the Judicial Field”. By 2025 courts were to get an AI support system, and by 2030 rules for the whole process.

The limits are written in the same document. The use of AI must not harm national security, reveal state secrets or break the security of personal data, and judges always make the decisions.

AI results serve as an extra reference in judicial oversight and management. The court names the goal directly: to take small routine work off judges and make legal resources easier for people to reach.

AI is also meant to watch the process: as the court plans it, the system will detect violations in case handling automatically.

The base for this is already built. According to the court, in 2021 Chinese courts accepted more than 11.43 million cases online, and by September 2022 more than 90,000 mediation organizations and more than 350,000 mediators had joined the online dispute resolution platform.

England and Wales

England regulates AI in courts with guidance for judges, not with a law. On 31 October 2025 the judiciary of England and Wales issued an updated version to replace the April 2025 one.

The guidance covers all judges, their clerks, assistants, legal advisers and court staff. The lead judge for AI, Lord Justice Colin Birss, reminded that judges’ use of AI must be consistent with the duty to protect the integrity of justice.

Three rules of the guidance: don’t enter private information into public AI services, keep in mind the risk of invented answers, and take personal responsibility for material issued in your name.

The guidance separates tasks. It calls summarizing large texts, emails and administrative work potentially useful, and it does not recommend public chatbots for searching new information that cannot be checked, or for legal analysis.

There are new threats too. The guidance warns about hidden text (white text) that the system sees but the reader does not, and requires work with AI on official devices, not personal ones.

For lawyers the conclusion is direct. A representative is responsible for the accuracy of what they file, and a judge may ask for confirmation that the lawyer personally checked citations obtained with a chatbot.

Colombia: the first ruling with ChatGPT

On 31 January 2023 Judge Juan Manuel Padilla (Cartagena) included ChatGPT answers in a ruling. They made up about 29% of the text: 2 pages out of 7, with four chatbot answers.

The case was simple: an insurance company demanded an extra payment for a medical procedure for a child with autism spectrum disorder. The appeal judge upheld the ruling in the child’s favor and used ChatGPT for the reasoning.

Ten days later, magistrate Maria Victoria Quinones of the Administrative Tribunal of Magdalena also cited ChatGPT prompts, in an order on a hearing held in the metaverse.

The judges did not hide ChatGPT and put its text in quotation marks. Researcher Juan David Gutierrez, writing on Verfassungsblog, still called this a mistake: in his view, the judges treated the chatbot as an oracle whose answer needs no checking.

Colombia’s Constitutional Court gave its answer in ruling T-323 of 2024. AI may be used in administrative and document work, as support for the judge, and to correct and shorten texts under a lawyer’s supervision, but it must not replace the judge’s logical and human reasoning.

On 26 September 2024 the Superior Council of the Judiciary issued circular PCSJC2437 with a summary of this ruling for the country’s courts.

EU: high risk

The EU AI Act, in Annex III, point 8(a), classes as high risk the systems that help courts or alternative dispute resolution bodies research facts and apply the law.

In its guidance on this point, the European Commission draws the line by function. A system that prepares a draft ruling from the facts of the case, the parties’ positions and case law is high risk.

Assigning cases to judges by specialization and workload, anonymizing rulings, and checking fee payments and representatives’ authority count as supporting administrative work. It does not affect how a specific case is decided and is not high risk.

The point covers more than state courts: it names alternative dispute resolution bodies directly. So firms that handle disputes involving European parties should take it into account.

The line can move. The assessment of case assignment may change if the system itself evaluates a specific case, while a service that only explains procedures, deadlines and documents to citizens does not fall under point 8(a).

The deadlines have moved. The Digital Omnibus package moved the obligations for Annex III high-risk systems from 2 August 2026 to 2 December 2027; the amending regulation took effect on 27 July 2026.

Russia: the 2026 concept

Order No. 59-P of the Supreme Court of Russia of 3 September 2026 approved the Concept for introducing AI technologies into court proceedings: recording hearings, assigning cases, preparing draft rulings. AI stays a tool of the judge, and a person makes the decision.

Under the schedule, all courts get access to judicial AI services in Q2 2027. By 2030 more than 95% of judges should use them regularly, and the time to hear simple cases should fall by 20–30%.

The path to these numbers goes in stages: each solution passes a process audit, a pilot and an assessment of results, and only then scaling. The Center for Judicial Competencies at the Supreme Court sums up the pilots and updates the risk matrix at least every six months.

In scale and timing, the Russian model is closer to China’s. In the split of roles it matches England and Colombia: AI prepares the material, and the judge is responsible for the decision. On top of that, the Concept gives a party the right to know that AI was used and to challenge its conclusions.

Unlike the EU, the Concept does not divide functions by risk level. Case assignment and draft rulings sit on one list, and protection comes from the risk matrix, the judge’s check of the result and a ban on public AI services.

All five systems have one thing in common: AI speeds up the preparation of materials, while the decision and the responsibility stay with a person. For a firm this is a ready standard: AI prepares a draft and finds case law, and the lawyer checks the citations and signs.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. China Daily, 12.12.2022 (govt.chinadaily.com.cn)
  2. Courts and Tribunals Judiciary, 31.10.2025 (judiciary.uk)
  3. Verfassungsblog, 23.02.2023 (verfassungsblog.de)
  4. AI Act Service Desk, European Commission, 2024 (ai-act-service-desk.ec.europa.eu)
  5. Usercentrics, 07.2026 (usercentrics.com)
  6. ppt.ru; Garant, September 2026 (ppt.ru)
  7. Courts and Tribunals Judiciary, AI Judicial Guidance, 31.10.2025 (judiciary.uk)
  8. Rama Judicial de Colombia, Consejo Superior de la Judicatura, 26.09.2024 (ramajudicial.gov.co)
  9. Garant, text of the order, 3 September 2026 (base.garant.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us