Introduction
The growing use of artificial intelligence in criminal investigations is transforming not only the way evidence is analysed, but also the conditions under which procedural rights can be exercised. As algorithmic systems become increasingly involved in the collection, processing and interpretation of information, new questions emerge regarding transparency, comprehensibility, human oversight and accountability.
The DIGITAL RIGHTS project explores these challenges and examines the safeguards necessary to ensure that technological innovation remains compatible with the fundamental principles of fair trial rights.
Understanding Algorithmic Evidence in Criminal Proceedings
Impact of the Digitisation of Criminal Proceedings
The digitisation of criminal proceedings is not limited to the digitisation of procedures. It brings about a more profound transformation: that of the way evidence is gathered.
In contemporary investigations, digital evidence is no longer merely collected. It is generated by algorithmic analysis systems capable of structuring data, identifying correlations and guiding investigative hypotheses. Evidence thus becomes partly dependent on complex technical operations, the logic of which is not immediately apparent to those involved in the proceedings.
This development gives rise to a major tension: can the adversarial process continue without the evidence being comprehensible?
The DIGITAL RIGHTS project highlights a risk of structural opacity. When evidential findings are based on tools whose parameters, models or internal processes are beyond the parties’ understanding, the focus of the discussion regarding the evidence shifts. It no longer centres on the content, but on the trust placed in the tool.
However, criminal proceedings cannot be based on a logic of technical trust. They presuppose the possibility of a genuine debate, which implies that the evidence presented must be comprehensible, explainable, and open to challenge.
Right to Transparency
Algorithmic transparency must therefore be recognised as a prerequisite for the right to be heard, rather than merely a regulatory requirement. It entails access to the processing logic, the data used and the analysis parameters.
The EU Regulation on artificial intelligence marks an important milestone by classifying systems used in criminal matters as ‘high-risk’. However, this risk-based approach does not obviate the need for procedural consideration. The question remains as to how this transparency can be translated into effective legal rights for the defence.
Ultimately, the digitisation of criminal proceedings requires a shift in the focus of safeguards. It is no longer just the evidence that must be accessible, but the way in which it is produced.
Human Oversight and Accountability in AI-Assisted Investigations
From Transparency to Human Oversight
The requirement that algorithmic evidence be understandable and contestable naturally leads to a broader question: how can procedural fairness be preserved when investigative processes increasingly rely on artificial intelligence systems?
Beyond transparency, attention must also be paid to the mechanisms through which human actors retain control over AI-assisted decision-making and remain accountable for the use of algorithmic systems.
AI-Assisted Tools in Criminal Investigations
The deployment of AI-assisted tools in criminal investigations introduces a fundamental shift in the architecture of decision-making. Investigative processes are increasingly supported — and, in some cases, guided — by algorithmic systems capable of processing complex datasets and generating analytical outputs at scale.
This evolution raises a central question of governance: where does responsibility lie in an environment where human decisions are shaped by machine-generated insights?
The DIGITAL RIGHTS project identifies the risk of a progressive “normalisation” of algorithmic authority. As reliance on AI tools increases, there is a tendency for their outputs to acquire a presumption of reliability, potentially reducing the degree of critical scrutiny applied by human decision-makers.
Excessive reliance on algorithmic outputs may lead decision-makers to attribute a degree of objectivity and reliability that exceeds the actual capabilities of the technology. This phenomenon, often described as automation bias, may weaken the critical assessment that remains essential to procedural fairness.
Human Oversight Requirement
Human oversight must therefore be understood as more than a formal requirement. It must constitute a substantive capacity to interrogate, contextualise and, where necessary, reject algorithmic outputs. This implies that practitioners are not merely users of AI systems, but informed actors capable of engaging with their limitations.
The EU AI Act provides an essential regulatory anchor by requiring that high-risk systems remain subject to meaningful human control. However, the effectiveness of this requirement depends on its procedural translation. Oversight cannot be reduced to the existence of a human “in the loop”; it must ensure that human intervention is informed, autonomous and consequential.
Accountability and Traceability
Accountability, in turn, depends on traceability. Every use of an AI system in a criminal investigation should be accompanied by documentation that allows reconstruction of:
- the data inputs;
- the analytical processes;
- the role of the algorithm in decision-making.
Without such traceability, judicial review becomes structurally limited.
A further implication concerns the position of the defence. The right to challenge algorithmic evidence presupposes access not only to results, but to the conditions under which those results were produced. Courts must therefore recognise that contestability is a constitutive element of fairness in AI-assisted proceedings.
Ultimately, the integration of AI into criminal justice systems tests the resilience of fundamental procedural principles. If left unchecked, it risks introducing a form of decision-making that is technically efficient but procedurally opaque.
Maintaining the primacy of human judgment is not a conservative reflex; it is a normative requirement.
Quick Takeaway
Core challenges related to AI-assisted investigations:
- Algorithmic opacity
- Limited transparency of analytical processes
- Risk of excessive reliance on automated outputs
- Insufficient human oversight
- Lack of traceability and accountability
Ensuring procedural fairness requires both comprehensible algorithmic evidence and meaningful human control throughout criminal proceedings.
Effective Procedural Safeguards in the Age of AI
Taken together, these reflections demonstrate that procedural fairness in AI-assisted criminal proceedings depends on two complementary conditions.
First, algorithmic evidence must be sufficiently transparent and comprehensible to enable meaningful scrutiny and challenge by all parties involved in the proceedings.
Second, the use of artificial intelligence systems must remain subject to genuine human oversight, supported by mechanisms of accountability and traceability that preserve responsibility for decisions affecting individuals.
Key Insight
The protection of procedural rights in AI-assisted proceedings requires more than access to algorithmic outputs. It also depends on the ability to understand how those outputs are produced, to challenge their reliability, and to ensure that human actors remain responsible for their use throughout the criminal justice process.
Looking Ahead: Cross-Border Digital Investigations and Judicial Cooperation
The challenges associated with AI-assisted investigations become even more complex when criminal proceedings extend beyond national borders. The exchange of electronic evidence, the use of digital cooperation tools and the application of different procedural frameworks raise additional questions regarding fairness, transparency and defence rights across jurisdictions.
Read next: Cross-Border Digital Investigations and Judicial Cooperation
Français