Artificial intelligence in court: How the judiciary is digitally rearming
Automated AI software for judgment anonymization is running in live operation, but the final decision in court must be made by a human.
(Image: Summit Art Creations/Shutterstock.com)
The digitalization of the judiciary is gaining momentum. However, when it comes to artificial intelligence, expectations and statistical certainty diverge. In judicial circles, there are increasing complaints that algorithmic text generators are flooding the courts with automated filings. However, the federal government lacks evidence for this. In its response to an inquiry from the AfD parliamentary group, it states that it has no concrete findings on a current increase in AI-generated applications. There are no systematic surveys or reliable figures on this.
It is known that AI systems are also used in everyday legal practice for drafting submissions, writes the lead Federal Ministry of Justice. However, this is not an entirely new development.
At the same time, the state is advancing the use of AI on the other side of the judge's bench. There are numerous projects in the German judicial landscape where algorithms are being tested or have already been partially put into live operation. According to the information, the goal is to relieve the judiciary in dealing with its backlog of cases and to optimize processes. The focus is on automating uniform, standardized work steps. These include, in particular, the automated anonymization of judgments, the structuring of procedural documents, and the evaluation and extraction of relevant data.
Field test in the federal states
Various state projects show what this looks like in practice. Under the names “Aleks” in Lower Saxony and Bavaria, as well as “Jano” in Hesse and Baden-Württemberg, relevant AI applications for anonymization are running. The latter went into regular operation in civil jurisdiction in December.
The experience so far is said to be positive, as the algorithms have already achieved higher hit rates in routine tasks than manual processing. Another example is the Lower Saxony project “Maki”. It is designed to facilitate the handling of mass proceedings and to support the legal analysis of files and submissions.
A current case shows that concerns about unchecked AI content in legal transactions are not unfounded. Fabricated case numbers and completely hallucinated judgments in appeal submissions caused trouble. Berlin judges therefore felt compelled to reprimand the bar.
European guardrails
The federal government is aware of the risks associated with AI-supported decisions. It warns of distortions due to incorrect or incomplete data, algorithmic illusions, and the human tendency to blindly trust automated suggestions. Lack of traceability and susceptibility to manipulation are also issues. Regulatory guardrails are in place to counter these dangers.
The use of AI in judicial and prosecutorial proceedings is primarily governed by the EU AI Regulation, according to the Ministry of Justice. This consistently classifies such decision-support systems as high-risk AI. This entails legal obligations regarding risk management, data quality, complete documentation, cybersecurity, and human supervision.
In parallel, the federal and state governments adopted a joint AI strategy for the judiciary last year to ensure legally sound and uniform application. Standards for risk management systems and binding guidelines on data governance and quality are currently being developed within this framework.
Human responsibility
Above the technological developments stands a dogma of the German Basic Law: the constitutional framework of judicial independence sets limits to automation. Judicial power is exclusively reserved for elected judges, the ministry emphasizes. The final decision in a legal dispute must therefore always be made by a human and personally accounted for. AI may only act in a supporting role. Human supervision serves as a control mechanism to identify and correct machine errors in a timely manner.
To ensure that judicial staff are capable of exercising this function, the executive branch relies on targeted training. Competencies regarding the technical, legal, and ethical issues surrounding the technology are to be strengthened through new initial and continuing education offerings. Part of the national AI strategy is the creation of a competence framework to establish a common understanding and derive tailored training measures.
Millions for the digital pillar of the rule of law
The transformation is financed and driven by a cooperation between the federal government and the states. The strategy adopted in the summer of 2025 forms the foundation, which, according to current plans, should be largely implemented by the end of 2026. The federal government is participating financially and technically in the states' AI projects as part of its digitalization initiative. This seed funding also runs until December.
After that, the new Pact for the Rule of Law is to follow seamlessly. Its so-called digital pillar provides for further funding with a term of three years. Whether and to what extent there is a need for further legislative action in the area of European data protection law or in liability issues for automated processing is still being examined by the executive branch. (nen)
