The Russian judicial system is taking an ambitious step into the digital future. As became known from my sources, pilot testing has begun on four large language models (LLMs) at once, which are being trained on real judicial practice. This is not just an experiment, but a systematic effort to integrate artificial intelligence into justice, as was announced at the BRICS Supreme Court Presidents' Forum in New Delhi.
What is really happening?
The models are not just analyzing texts—they are mastering the legal positions of the Supreme Court and key rulings that affect the lives of citizens and businesses. Testing is underway in the regions, and the goal is not to replace a judge, but to create a tool for improving the quality of justice.
Judges, as emphasized, have already shown high interest in the technology. They see in it the potential to unify practice, eliminate logical contradictions, and reduce the number of errors. However, implementation is proceeding with extreme caution. A concept has been approved that clearly defines the tasks and principles of AI operation in judicial proceedings. A specially created Center for Judicial Competencies in the Field of AI is responsible for implementation, which will collect best practices and interact with the scientific and IT community.
Ethics and the limits of what is permissible
Special emphasis is placed on ethics. Together with leading developers, principles for the use of neural networks have been formulated. Judges are undergoing training to understand how the models work and where the line lies between assistance and interference. Students have been brought into the work—an all-Russian competition is being held to create AI prototypes for justice.
The role of the algorithm is also strictly defined: it is an assistant, but not a subject of law. The evaluation of evidence and decision-making are the prerogative of a human being, based on the law and inner conviction. A probabilistic approach must not replace professional judgment.
Issues of technological independence have also been raised at the international level. A proposal was made to develop common ethical principles for BRICS countries so that judicial infrastructure does not depend on external jurisdictions, and judges can exchange experience through internships.
This is only part of a large-scale course toward the algorithmization of public administration. The authorities are encouraging the transition to domestic software, and AI priorities in security and education are being discussed. By 2030, for example, it is planned to transfer half of the processes in construction and housing and communal services to AI.
My view: This is a landmark precedent. The crypto industry has long talked about verifiability and transparency, and here we see how the same principles are being attempted to apply to fiduciary systems. However, the key risk is not technical failures, but the algorithm's "black box." If a court cannot explain why the model produced a particular recommendation, this will undermine trust in justice. Success will depend on how well human control can be maintained and AI explainability ensured.