A systemic approach to the scientific support of the judicial system

AI is a tool to assist the judge — not a replacement

Eight fundamental principles of the Concept define the boundaries within which the use of artificial intelligence in judicial proceedings is permissible.

• 01

  Protection of human rights and freedoms

  No violation of constitutionally guaranteed rights and freedoms and no discrimination in the design of technological solutions.

• 02

  Auxiliary nature

  AI is a tool to assist the judge, not to replace them. The final decision always rests with a human being.

  // final say remains with the judge

• 03

  Legality

  AI is applied strictly within the framework of the legislation governing judicial proceedings.

• 04

  Transparency and explainability

  Algorithms must be understandable, verifiable and auditable; every interaction with an AI system is recorded.

• 05

  Quality and reliability

  Source data comes only from official sources; the outputs of AI systems are verified and reliable.

• 06

  Public order and ethics

  The use of AI does not conflict with public interests or ethical norms.

• 07

  Data confidentiality

  Strict compliance with requirements for the automated processing of personal data.

• 08

  Technological independence

  Predominant use of domestic technologies and solutions in the field of artificial intelligence.

Every decision of the Center is checked against all eight principles — before any pilot is launched.

End-to-end AI across the entire lifecycle of a case

Seven directions of the Concept — from filing documents to the enforcement of judicial acts and review by higher instances.

Every solution goes through a full innovation cycle

Five stages — from hypothesis to post-project monitoring. The decision to launch, continue, adjust or close a project is based on objective monitoring data.

• STAGE 01

  Initiation

  Identifying the problem, project hypothesis, expected results, success criteria and constraints.

• STAGE 02

  Planning

  Project passport: key milestones, working group composition, resources and partners, communications and risk plan.

• STAGE 03

  Execution

  Research, solution prototype, testing in pilot courts — within a regulatory sandbox where necessary.

• STAGE 04

  Completion

  Assessment against success criteria, final analytical report, decision: scale, refine or close.

• STAGE 05

  Monitoring

  Observation of the implemented solution, assessment of its stability and impact on judicial performance indicators.

• // evidence-based approach

  Decisions are made on the basis of objective monitoring and analysis data.

• // right to a negative result

  An unsuccessful pilot is an acceptable outcome, provided it is properly documented and lessons are drawn.

• // external cooperation

  Joint working groups and regulatory sandboxes with technology partners and academia.

• // transparency

  Project registry, regular reports to the Chairman of the Supreme Court, information for stakeholders.

Calendar plan

Annex No. 1 to the Concept. Control points — from forming the Center to giving all courts access to judicial AI services.

• Q3 · 2026

  Center of Judicial Competencies established

  The Center has begun implementing the Concept — Order No. 60-P of 3 September 2026.

  Supreme Court

• Q3 · 2026

  Strategic session “Applied use of AI technologies”

  Judicial system staff briefed on the current state of AI technologies and promising directions for their introduction.

  Supreme Court

• Q3 · 2026

  Results of pilot use of AI products for official tasks

  Parameters defined for the software, infrastructure and equipment needed to automate official processes.

  Judicial Department Supreme Court

• Q4 · 2026

  Decision on priority technological solutions

  Specific software, deployment procedure for AI services in the closed circuit of the judicial system, funding sources.

  Judicial Department Supreme Court

• Q4 · 2026

  Priority scenarios for judicial AI services

  Priority scenarios for the official use of AI defined.

  Supreme Court Judicial Department

• Q1 · 2027

  Pilot of priority scenarios

  Scaling procedure for the most in-demand scenarios and judge training on the specifics of AI tools.

  Judicial Department Supreme Court · RSUJ

• Q2 · 2027

  All courts get access to judicial AI services

  AI services used by judges in their work on a permanent basis.

  Judicial Department Supreme Court

• Q2 · 2027

  Pilot of additional AI usage scenarios

  Additional AI tools for official tasks implemented.

  Judicial Department Supreme Court

Measurable targets by 2030

Annex No. 2 to the Concept. Nine quantitative indicators against which the implementation results will be measured.

• >95%

  Share of judges using judicial AI services on a permanent basis

• 20–30%

  Reduction of consideration time for simple (standard) cases

• ≥ 90%

  Share of court hearing records prepared with AI

• 30–50%

  Share of AI-generated drafts of procedural documents

• ≥ 80%

  Share of drafts approved by judges without significant edits

• ≥ 90%

  Accuracy of evidence analysis and contradiction detection

• ≤ 3 sec

  Time for a judge to find relevant judicial acts

• 15–20%

  Reduction in the number of complaints and modifications (reversals) of judicial acts

• ≥ 30%

  Share of process participants using judicial AI services

Implementation risk matrix

Annex No. 3 to the Concept. Updated by the Center as the measures are implemented — at least once every six months.

• Model hallucinations — non-existent norms, acts, citations

  Likelihood: high. Impact: very high. Mitigation: Mandatory RAG architecture with source references, verification by the judge, ban on public AI services, model stress-testing.

• Systematic bias in training data

  Likelihood: medium. Impact: high. Mitigation: Audit of training corpora, bias tests on protected attributes, monitoring of deviations by region and case category.

• Leakage of personal data and restricted-access information

  Likelihood: medium. Impact: very high. Mitigation: Isolated circuit, user training, ban on transferring data to public clouds.

• Cyberattacks on AI infrastructure

  Likelihood: high. Impact: very high. Mitigation: Redundancy, domestic hardware, regular penetration tests, cooperation with supervisory and law-enforcement bodies.

• Reduced independence and initiative of judges

  Likelihood: medium. Impact: high. Mitigation: Ban on “default choice”, AI positioned solely as an assistant, mandatory reasoning when rejecting an AI recommendation, interaction logs.

• Profiling of judges by their decisions

  Likelihood: low. Impact: high. Mitigation: Complete ban on personalized analysis of judges, anonymization of judge-related data in analytics.

• Technological dependence on foreign models

  Likelihood: medium. Impact: high. Mitigation: Domestic technological solutions, mandatory deployment in the closed circuit.

• Resistance from the judicial community

  Likelihood: medium. Impact: medium. Mitigation: Involving judges in scenario design, training, gradual feature rollout, voluntary use of most features during the pilot stage.

• Model quality degradation over time

  Likelihood: high. Impact: medium. Mitigation: Continuous quality monitoring, periodic retraining, A/B testing protocol for new versions.

• AI-generated materials with inaccurate references submitted by parties

  Likelihood: medium. Impact: high. Mitigation: Mandatory disclosure of AI use when filing documents, automatic verification of citations against acts and regulations, procedural sanctions.

• Budget risks — funding disruptions at individual stages

  Likelihood: medium. Impact: high. Mitigation: Multi-channel funding, prioritization of quick-return (ROI) directions.

• Violation of the parties' right to a fair trial

  Likelihood: low. Impact: very high. Mitigation: The party's right to know that AI is used, the right to challenge AI findings, the judge's mandatory final role.

The matrix is updated at least once every six months. The Head of the Center bears personal responsibility for the results of its work and their scientific validity.

Head, four deputies, coordinator and experts

The composition is determined by the Chairman of the Supreme Court of the Russian Federation. The Center engages staff of the Supreme Court apparatus, the Judicial Department, the Russian State University of Justice named after V.M. Lebedev, judges of lower courts and, where necessary, representatives of government bodies, academia and technology partners.

• head

  D.A. Kunev

  Justice of the Supreme Court of the Russian Federation

• deputy

  A.A. Simonov

  Head of the Department for Systematization of Legislation and Analysis of Judicial Practice, Supreme Court

• deputy

  D.O. Antonov

  Head of the Department of Informatization and Communications, Supreme Court

• deputy

  S.A. Besschastny

  Deputy Director General of the Judicial Department at the Supreme Court

• deputy

  A.V. Lisiev

  Deputy Director General of the Judicial Department at the Supreme Court

• coordinatorexperts

  Coordinator and experts

  Cycle planning, cooperation management, personal responsibility for the quality and reliability of project results

Regulatory framework

Founding documents of the Center — orders of the Chairman of the Supreme Court of the Russian Federation (in Russian).

Building digital justice together