Ancillary function
Document search, conversion, duplicate detection and filing support without influencing legal analysis.
Measures 4.02 to 4.05
Artificial Intelligence in Justice: Uses and Safeguards
Artificial intelligence may support judicial tasks, but controls must increase as a tool approaches legal reasoning and decision-making.
Four proposed uses
Measures 4.02 to 4.05 cover pre-processing of simple files, case-law research, consistency analysis by sampling and broader deployment in courts. These functions do not carry the same risk. Document classification is different from a system that influences the outcome of a case.
Core distinction
Reasoning support
Targeted case-law research, comparison and draft summaries. Enhanced control is required.
Determinative influence
Scores or recommendations likely to orient the outcome. Highest safeguards apply.
The EU AI Act
Regulation 2024/1689 classifies certain systems used by or on behalf of judicial authorities to assist in researching and interpreting facts and law, or applying law to concrete facts, as high-risk. Purely ancillary administrative activities are distinguished.
High-risk obligations include risk management, data governance, documentation, logging, transparency, human oversight, accuracy, robustness and cybersecurity.
Eight questions before deployment
- Purpose: what concrete problem is being solved?
- Legal basis: what authorises the processing and use?
- Risk classification: ancillary, decision support or determinative influence?
- Data: are they lawful, relevant and documented?
- Performance: how are error, bias and drift measured?
- Human control: who reviews and with what competence?
- Audit trail: which versions, inputs and outputs are kept?
- Remedy: how can a person challenge the process?
Working classification of uses
| Use | Working position | Main condition |
|---|---|---|
| Case-law research | Potentially authorised | Verifiable sources and traceable citations. |
| Document classification | Gradual deployment | Quality control and correction mechanisms. |
| File summary | Controlled experiment | Systematic verification against original documents. |
| Consistency analysis | Experiment and audit | No secret scoring or artificial uniformity. |
| Draft decision | High risk | Explicit framework, enhanced oversight and no autonomous final outcome. |
| Autonomous judgment | Excluded | Judging remains a human function. |
Productivity must be measured net
Net gain = time saved − review time − operating costs − error costs
Downloads
Caution: GDPR, law-enforcement data rules, national data-protection law and the AI Act must be mapped for each use, court and data category.