Why Contestability Matters More Than Admissibility in Criminal Procedure

Consider a prosecution that turns on messages said to have been taken from an encrypted communications platform. The messages place the accused at the centre of an organised operation, and the court has ruled them admissible. The defence then asks the obvious questions. How were the messages captured? How were they attributed to a particular handset? What processing did they pass through before they reached the case file, and was anything lost or changed along the way? The answers come back incomplete, or withheld on grounds of confidentiality, or simply unavailable because the relevant knowledge sits with a foreign authority or a private operator. The evidence is in the trial and it weighs heavily, yet there is no real way to examine it.

This is no longer an unusual situation. Encrypted intercepts, automated analytics and machine-assisted identifications now appear regularly in serious cases, and they behave differently from the evidence our procedure was designed around. A witness can be cross-examined. A document has a chain of custody that can be reconstructed. Evidence of this newer kind often cannot be opened up at all, because the method that produced it is protected, distributed across jurisdictions, or opaque even to the people who operate it.

Criminal procedure has a familiar device for controlling what enters a trial. Admissibility asks whether a given piece of evidence may be used, given the rules on relevance, lawful origin and exclusion. It is a threshold judgment, made once, and it is generally satisfied as soon as the lawful origin of the evidence has been established. What it does not ask is whether the party against whom the evidence is used can do anything with it afterwards. For most of the history of the trial that omission caused little difficulty, since evidence that was admitted could ordinarily be tested in the normal course of things. Opaque evidence breaks that quiet assumption. The gate opens, and what lies beyond it stays in shadow.

I have argued elsewhere, in work that traces the move from admissibility to contestability, that the safeguard which actually matters here is the practical capacity of the defence to challenge evidence throughout its life. That capacity reaches the source of the evidence, the manner of its collection, the processing it underwent, its reliability, the way it has been interpreted, what has been disclosed about it, and the use an institution finally makes of it. Admissibility and contestability are not the same question asked twice. One concerns entry. The other concerns whether, once the evidence is inside, the accused has any genuine purchase on it. A judge may be entirely correct that material was lawfully obtained while the defence is left with no way to interrogate how it was generated, and in the hardest cases that is exactly what happens.

Contestation that means something has fairly concrete requirements. The defence needs to know enough about a method to find the places where it might fail, whether that is the assumptions built into a model, the conditions under which an intercept was taken, or the error characteristics of an automated process. It needs time, expertise, and often an independent route to the underlying material. It needs a setting in which raising these points can actually affect the verdict rather than merely being recorded. This is what equality of arms and the adversarial principle protected by Article 6 of the European Convention on Human Rights come to in practice. A right to object on paper, unaccompanied by any real opportunity to test the case, satisfies neither.

When the method behind a piece of evidence is shielded by confidentiality, foreign sovereignty or commercial secrecy, the system faces a choice it cannot avoid. It can find a controlled way to make the protected material examinable, or it can reduce the weight the evidence is allowed to carry. Letting such evidence count at full strength while denying any means of testing it keeps the adversarial form and empties it of content.

Regulation has started to approach the problem from the side of the technology. The European Union's AI Act places systems used in the administration of justice in its high-risk category and attaches duties of transparency, human oversight and documentation to them. The Council of Europe's ethical charter on artificial intelligence in judicial systems sets out comparable expectations around quality, security and user control. Instruments of that sort are useful, and they discipline how tools are built and deployed. They do not resolve the position of one defendant confronting one piece of opaque evidence in one trial, which remains a matter for procedure.

The concern reaches past the evidential. Criminal procedure exists so that the accusatory power of the state stays answerable to the person it is used against. Admissibility guards the entrance. It was never meant to ensure that the accused can engage with whatever comes through. Where the decisive proof is generated by systems that neither side fully controls and only one side can reach, the imbalance stops being a matter of evidence and becomes a matter of institutional power, and a procedure that treats admissibility as its principal protection rests fairness on a question the case no longer turns on.

None of this calls for new categories of evidence or for abandoning the admissibility rules courts already apply. It asks for something smaller and harder. Treat admission as the beginning of the inquiry into fairness rather than its end, and put a second question to opaque evidence. Can it be contested? Where the honest answer is that it cannot, that answer should bear on whether the evidence is allowed to count, and on how much. A trial that cannot be challenged on its most important proof has quietly ceased to be adversarial in the way the principle was meant to secure.