AI in the Courtroom: AI Prosecutions vs Defence Rights
16 July 2026 | AI, Criminal Law, Data Protection, Human Rights
Opinion piece by Professor Stefano Filletti
Introduction
The battleground for justice has fundamentally, and irrevocably, shifted. For centuries, criminal law orbited around the physical: the locus delicti, the tangible weapon, the paper list, the witness in the box. Justice, while difficult, was tactile. Today, it orbits around the ethereal: the data packet, the encrypted server, the blockchain ledger, the digital footprint.
As I've argued in previous articles evidence is no longer just in the room; it is the room, the cloud, the network. The digital environment is not merely a source of evidence; it is the venue of the crime itself.
And this new paradigm, this digital deluge, presents the single greatest challenge to the most sacrosanct principles of our shared legal heritage: the fundamental rights enshrined in our Constitutions, in the European Convention, and in the Charter. I speak, of course, of the right to a fair trial.
My presentation today is not one of technological fear. It is not a plea to return to a simpler time. It is, instead, a call of procedural urgency. Question is: are we safeguarding fair trials beyond in this digital age? The truth is, the erosion is already happening. The foundations are already cracked.
Our central problem is this: the speed of digital investigation, championed by powerful new tools and instruments of mutual recognition, is catastrophically outpacing the process of justice. The law of technology, dictates exponential change. But the law of justice, due process, demands deliberation. It demands time. It demands scrutiny.
And in the gap that forms between this technological speed and this procedural deliberation—a gap of time, resources, and law—defence rights are being extinguished. Not by malice, but by design. By an architectural flaw in our new system of justice. A system that has prioritised efficiency over equity.
The New Paradigm: Data as the New Dagger
We are in an era of 'data-driven' prosecution. And this is logical. Look at our agenda. The EPPO, for instance, was established to combat complex, cross-border VAT fraud and organised crime. These are crimes that proliferate due to the digital networks that facilitate them.
To fight this, we have armed our States and our EU bodies with powerful, fast instruments. The European Investigation Order (EIO) is the crown jewel of this new arsenal. It is a brilliant piece of legal engineering, designed for speed and built on the sacred principle of mutual recognition. It is designed to cut through the bureaucratic fog of old MLA requests.
We see this same emphasis on speed in technology. We see AI-driven forensics that can sift terabytes—petabytes—of data in hours, not years. We see collaborative, multi-state operations, that can dismantle a criminal network in a single morning.
This is all necessary for the investigation of crime. But I ask you to pause and consider: what is its consequence for the defence of the accused?
The consequence is a profound, structural imbalance. The State, including supranational bodies like the EPPO, now wields the power of instantaneous, large-scale, cross-border data seizure.
But the rights of the defendant—the right to disclosure, the right to prepare a defence, the equality of arms, the right to confidentiality with one's lawyer—remain tied to analog-era procedures. They are slow, they are resource-intensive, and they are increasingly being rendered obsolete.
The very premise of mutual recognition—that we trust each other's systems—is being strained to the breaking point. We are no longer asking a State to hand over a specific file or a bank statement. We are asking it to hand over an entire server. An entire digital life. And we are presuming that the rights of the individual, caught in that digital dragnet, will be protected.
The volume of data seized is, in itself, a new form of procedural barrier. When the prosecution serves a hard drive containing 5 terabytes of data on the defence, is that disclosure? Or is that a 'Digital Fishing Expedition' masquerading as disclosure? Handing someone a library and telling them the evidence is "in there" is not disclosure. It is an abdication of the prosecutor's duty to particularise the case.
This is not a theoretical 'problematica.' It is a clear and present danger. And it is amplified by the gross asymmetry of resources. The State has forensic labs, teams of technicians, and sophisticated AI platforms. The defence, very often, has a standard laptop and a commercial PDF reader.
This asymmetry is where the ideal of 'equality of arms' breaks down. And I wish to illustrate this today with a recent, stark example from our own jurisdiction here in Malta: the Bharwani case.
The Case Study: The 'Bharwani' Paradox
The Bharwani case, on which I have also had occasion to write, is a perfect storm of our new digital reality. It is not an obscure academic hypothetical; it is the lived experience of our courts.
It involved an EPPO-led operation into sophisticated, cross-border VAT fraud. The Maltese authorities were required to execute two instruments against an individual simultaneously: a European Arrest Warrant for his surrender to Sweden, and a European Investigation Order for the seizure of his data.
Here is the procedural paradox that our courts had to face:
1. The EIO was executed. In accordance with its principles of speed and efficiency, the Maltese authorities seized the entirety of the defendant's digital footprint—his servers, his computers, his hard drives. And, as required, they immediately dispatched this trove of data to the requesting State.
2. Almost in the same breath, the defendant was brought before the Maltese Courts to face the EAW proceedings—to argue why he should not be surrendered.
Now, I ask you as jurists, as judges, as prosecutors, as defence counsel: how does one defend oneself in such a situation?
The legal basis for challenging an EAW is narrow, but it exists. You might argue that the warrant is flawed, that the facts alleged do not constitute dual criminality, or that surrender would breach your fundamental rights under Article 6 or Article 8 of the ECHR.
But to make any of those arguments, you need evidence. You need your data. The data that might contain the exculpatory proof. The data that might prove the locus delicti was elsewhere. The data that might explain the very transactions the EPPO alleges are fraudulent.
Where was that data? It was gone. It was in the hands of the other State that was, at that very moment, demanding his surrender.
The defendant, in this case, was left, quite literally, defenceless. He could not access his own data to challenge the warrant. The court was placed in an impossible position.
This is not just a 'problem'. This is a direct and flagrant violation of Article 6(3)(b) of the European Convention: the right to "adequate time and facilities for the preparation of his defence." What 'facilities' are adequate when the State has seized all of them?
Furthermore, it is an annihilation of the 'equality of arms' under Article 6(1). The State had all of the evidence. The defence had none.
And we must be clear: the EAW proceeding in the executing State—Malta, in this instance—is the only time and place the defendant can challenge his surrender. Once that plane takes off, the jurisdiction of the Maltese court is extinguished. By allowing the EIO to preempt the EAW, the system de facto obliterated the only forum for the right of defence to be exercised.
This is a procedural black hole. It is a situation where one EU instrument, the EIO, functionally annihilatesthe fundamental rights guaranteed under another, the EAW framework. It is, in fact, a denial of justice.
The Bharwani case is not an anomaly. It is a warning. It is the logical, structural, inevitable outcome of a system that prioritises investigative speed over procedural fairness. We have created a mechanism for data seizure that is so efficient, it pre-empts the very possibility of a defence.
The 'Sealed Box': Privilege in a World of Big Data
But the Bharwani case only reveals the first layer of the problem—the problem of access to evidence. The second, and perhaps more insidious, layer is the problem of content.
This brings me to the core issue of privileged information.
In the analog world, the protection of privilege was a physical, tactile process. If the police searched a lawyer's office, or a suspect's home, a lawyer could stand present. An officer would pick up a file. Counsel could assert privilege over that specific document. It could be sealed, given to the court, and reviewed in camera by a judge. The 'sealed box' was respected.
Today, the 'search' is not a search. It is a total data extraction. An entire server is imaged. A 'bit-for-bit' copy of a hard drive is made. Terabytes of data, spanning years, are taken in an instant. This is not surgical; it is indiscriminate.
And buried deep within that digital mountain is the 'sealed box' of Legal Professional Privilege. Emails to lawyers. Drafts of defence arguments. Confidential advice on the very matter being investigated. Internal memos discussing legal strategy.
Legal Professional Privilege is not a mere procedural nicety. The European Court of Human Rights has repeatedly affirmed, that lawyer-client confidentiality is a cornerstone of a fair trial under Article 6. It is not a right of the lawyer; it is a fundamental right of the client. It is the foundation upon which the entire defence is built.
The question for us, 'beyond 2025,' is this: How do we protect that privileged box when the State has seized the entire mountain?
Our old methods are failing. 'Taint teams' or 'filter teams'—where prosecutors not on the case are asked to filter the data—are, with all due respect, a procedural farce. It is the State policing itself. Who are these individuals? What are their qualifications? How can a defence team ever be certain that the privileged material, once seen, has not infected the prosecutorial mindset?
We are told that keyword searches will suffice. But this is hopelessly naive. Who sets the keywords? The prosecution. What if the privileged document does not contain the keyword 'lawyer' but 'my confidant'? What about encrypted files or messages? What if the communication is a 'memo to self' about a phone call with counsel? This method is designed to fail.
And now, as our agenda highlights, we are layering AI on top of this. The State is using algorithms to sift this data. Do we, as defence lawyers, have any insight into those algorithms? Is an AI capable of understanding the human nuance of legal privilege? Or is it trained by the State to find incriminating data, with privilege as an afterthought? This is a 'black box' review of evidence, which is anathema to the principle of inter partes procedure.
If the State sees the defence's hand, the trial is over before it begins. The equality of arms is not just broken; it is a legal fiction. You cannot have a fair trial when one side has been reading the other side's notes. This is not justice; it is surveillance.
And this "total data" seizure has a chilling effect far beyond legal privilege. In that terabyte of data is the suspect's entire life: medical records with his doctor, confidential communications with his spouse, perhaps even sensitive information from a journalistic source. The State's digital dragnet is violating all forms of privilege, creating a society where citisens must fear that all confidential communications can be seized and reviewed.
The Way Forward: Architecting the Safeguards
The challenges are profound. But they are not insurmountable. We are jurists. Our specialty is process.
The solution is not to abandon our new tools—the EPPO, the EIO, and AI are here to stay.
The solution is to re-infuse these tools with the fundamental rights they have left behind. The solution is to build the safeguards into the architecture of the system itself. We, as jurists, academics, and practitioners, have an onerous duty to architect these safeguards.
I propose four concrete, structural steps forward.
First, we must remedy the Bharwani paradox. We must legislate, at both the national and EU level, for a 'Digital Habeas Corpus.' This is not a new right; it is the modern application of Article 6(3)(b). It must be an automatic, fundamental right: if the State seises your digital life, you have an absolute right to a copy of that data. Crucially, this right must act as a stay on any related proceeding, like an EAW, until that disclosure is certified by a judge. And that copy must be in a searchable, accessible format—not a proprietary, unreadable forensic image.
Second, we must solve the "sealed box" problem. We can no longer rely on the prosecution to filter its own evidence. We must establish a new institution: an Independent Digital Magistrate or a judicially-supervised neutral technical expert. This body's sole function would be to receive the raw data, act as a true inter partes referee, allow the defence to propose its ownkeywords for privilege, and then segregate that privileged material before the prosecution or any other party reviews the data. This must be an adversarial process, not an ex parte one.
Third, we must redefine 'equality of arms' for the digital age. It must mean equality of digital arms. This means giving the defence not only access to the data, but the forensic tools and the time to analyze what may be millions of files. The EPPO and State prosecutors have vast forensic labs and budgets. The defence has a laptop. We must move towards State-funded, independent forensic experts for the defence, and access to the same forensic platforms (like Cellebrite or Encase) that the prosecution uses. This must be part of the legal aid framework. Without 'equality of resources,' there is no 'equality of arms.'
Finally, these rights need teeth. What happens when the State breaches these rules? We must champion a robust Digital Exclusionary Rule. If the prosecution accesses privileged material, or if data is transmitted in a Bharwani-style breach, the only effective remedy is the exclusion of that evidence and any evidence derived from it. This 'fruit of the poisonous tree' doctrine is the only deterrent that will ensure the State respects these fundamental procedural barriers.
The Valletta Protocol
As an acknowledgment to the shift to digitalisation in September 2025 the Valletta Protocol was signed. Officially the Third Additional Protocol to the 1959 COE Convention on Mutual Legal Assistance, it is not merely an incremental update but a strategic response to the realities of twenty-first-century crime. The Protocol was designed to directly address the challenge of cross-border access to electronic evidence, a critical function that complements our existing provisions on mutual legal assistance.
Its provisions are revolutionary for day-to-day judicial work. The Protocol champions the use of secure electronic channels for exchanging judicial information, from initial requests for assistance to the final delivery of evidence. Crucially, it allows for the direct transmission of requests between judicial authorities. This measure alone cuts through the bureaucratic fog that so often delays justice, enabling a prosecutor in one State to communicate with a judge in another with a speed and security that was once unimaginable.
Furthermore, the Protocol explicitly broadens the scope for using video conferencing for witness testimony and hearings. This seemingly simple tool has a profound impact, allowing for swift and efficient proceedings without the immense logistical and financial burden of cross-border travel. This is especially vital in cases involving vulnerable or at-risk witnesses. The Protocol also opens the door to more advanced technologies essential for modern investigations, such as data interception, digital transaction monitoring, big data processing, and even AI-driven forensics.
This having been said, the very speed and efficiency these tools offer can also pose a risk. In our haste to combat organised crime, we must never lose sight of the fundamental principles of due process and human rights that the Council of Europe was founded to protect. The ease of data sharing must not come at the cost of privacy. The speed of information exchange must not compromise a defendant's sacrosanct right to a fair trial.
This brings us to a crucial question: how can we work together to develop shared standards to regulate the use of these powerful tools? The answer, once again, lies in the spirit of cooperation embodied by the Valletta Protocol. It is not enough to simply use the same tools; we must agree on the same legal and ethical standards for their use.
Slide 12
First, we must establish common data and metadata standards. While the Protocol encourages electronic communication, we must ensure the data we exchange is in a format that is universally legible and forensically sound. We need to agree on what constitutes a "certified" digital document and how its integrity is maintained across borders.
Second, we must develop and implement robust security protocols. The systems we use to facilitate mutual legal assistance are a high-value target for organised crime and other malicious actors. We must invest collectively in state-of-the-art encryption and cybersecurity measures to ensure our judicial communications are impregnable.
Finally, and most crucially, we must invest in the people who use these tools. The Protocol and the digital tools it sanctions are only as effective as the legal practitioners who wield them. We must commit to continuous training and knowledge sharing, fostering a generation of judges, prosecutors, and lawyers who are not only digitally literate but who also deeply understand the legal and ethical ramifications of every digital action they take.
In conclusion, the digital age, and AI do not change our principles. They only test our commitment to them.
Technology is a tool. Justice is a principle.
AI is the new engine of justice. It is powerful, and it is fast. But engines without brakes and steering—which are due process and defence rights—do not lead to justice. They lead to collisions.
Our job, as jurists, as the collective legal mind of Europe, is to be the engineers of those brakes.
We are at a crossroads. If we continue to prioritise efficiency above all else, if we continue to accept procedural breaches as the 'cost of doing business' in a digital world, we will wake up in a system where the 'fair trial' is a nostalgic memory, a relic of an analog past.
Our goal must be to ensure that we do not sacrifice the process of the trial for the prosecsution of the crime.