Can AI Help Tackle the Criminal Court Backlog in England and Wales?
The Numbers
450,783 cases
The open caseload of the criminal courts of England and Wales at the end of March 2026[2]. But that is only part of the picture. The 80,061 cases open in the Crown Court alone have reached a record median age of 196 days, 22,124 have been open for at least one year, and contested jury cases have reached a median waiting time of 48 weeks.
In the Magistrate’s there were 77,877 trial-case receipts in Q1 2026, against 74,364 trial-case disposals. Receipts have exceeded disposals for trial cases in every quarter since Q1 2022.[3] The dilemma is clear: on top of an already enormous backlog, new cases are entering the system faster than the courts can dispose of them. The Crown Court did, however, for the first time in 3 years, dispose of more cases than it received in Q1 of 2026. Still, its backlog fell by only 37 cases, compared to a rise of 3,513 cases in the Magistrate’s.
Efficiency
So, therein lies the question, how can the criminal courts make a meaningful dent in their open caseload and start consistently seeing disposals exceed receipts? I am of the view that increasing the efficiency of the criminal courts via a meticulous and careful uptake of artificial intelligence may be a significant piece of the puzzle.
But, what do we mean by a more ‘efficient’ justice system. The Cambridge Dictionary defines efficiency first as ‘the quality of achieving the largest amount of useful work using as little energy, fuel, effort, etc. as possible’[4]. Indeed, when discussing an ‘efficient’ court system, cost-effectiveness and best utilising already limited resources is likely on our mind. However, as noted by Hedler, “the existence of other important institutional values gives rise to another type of efficiency: that of considering whether an action measures up to the values the organisation needs to uphold”[5]. In essence, we want to enable the criminal courts to work more expeditiously and reduce the enormous backlog, whilst not endangering due process and the rule of law.
The (Theoretical) Solution
It is no secret that our courts are somewhat… antiquated. To its credit, however, the Ministry of Justice (MoJ) has recognised that turning this metaphorical ship around will require significant changes to the way the justice system operates. The MoJ published its ‘AI Action Plan for Justice’ in July 2025, observing that “AI holds tremendous promise for addressing longstanding challenges in the justice system”[6] and the second of its three strategic priorities is to “embed AI across the justice system”.[7]
Investigatory Powers Commissioner, Sir Brian Leveson, went further in his Independent Review of the Criminal Courts Part II, suggesting that criminal processes should be redesigned around technology rather than simply automating inefficient existing processes, while insisting upon safeguards, transparency, ethical/legal standards and training[8].
It seems obvious, prima facie, that incorporating AI models to relieve administrative burdens across the justice system could free up enormous amounts of time and resources. One existing example comes from probation rather than the courts themselves. Justice Transcribe, an AI-powered transcription tool used by probation officers to help produce case notes, is estimated by the government to save the equivalent of 18,750 days of probation staff time annually.[9] This demonstrates the scale of administrative time that carefully targeted AI tools may be capable of releasing across the wider criminal justice system. The next question is whether AI could help reclaim some of that time for HMCTS staff, too.
Of the 7,939 cases scheduled for Q1 2026 in the Crown Court, 24% (~1,905) were ineffective (could not proceed as planned). Ineffective cases essentially mean duplicated workloads and costs. Advocates may need fresh preparation, the case must be relisted, witnesses must once more be prepared for trial and prisoners may need to be transported to the court again another day. Of the 3 primary factors contributing to ineffective trials, overlisting was by far the most significant, accounting for 27% (~514) of the total. Overlisting means more trials were scheduled to take place at a given court on a given day than that court had capacity for. Why does this happen? Because listing is unpredictable. Trials crack, advocates become unavailable, estimated trial durations are inaccurate etc. The current solution is, essentially, to list enough trials so that we are sure judges and advocates are never left twiddling their thumbs. We can surely do better?
Leveson has recommended the deployment of AI tools alongside a data-driven approach to listing in both the Crown and Magistrate’s courts[10]. If Google DeepMind’s GraphCast can be trained on decades of global weather data, outperforming a world leading predictive weather system in 90% of tested variables and producing 10-day forecasts in under a minute[11], it is not difficult to imagine predictive AI tools bringing greater predictive accuracy to court listing too.
I predict further improvements will also be seen upstream. In July 2026, the government accepted some major recommendations from Jonathan Fisher KC’s disclosure review. As such, PoliceAI will pilot AI analysis tools that can automatically generate summaries of digital evidence and will be supported with £75 million in government funding[12]. The potential impact of such a tool is significant. The government has stated that investigations can contain the equivalent of 500,000 ebooks of data and that the average fraud case contains more than four million documents.[13] Whilst it is obvious how this AI tool may improve the efficiency of police investigations, the indirect effects on criminal courts would be substantial also. 9% (~171) of the aforementioned ineffective cases in Q1 2026 were attributed to ‘prosecution not ready’.[14] If AI-tools can assist reviews and produce clearer and more comprehensive case materials, it is entirely plausible that those results will travel downstream to the CPS, enabling swifter decision making and case preparation, as well as reducing time spent by HMCTS staff in navigating poorly presented files. This, in theory, would lead to fewer adjournments and better use of limited courtroom capacity. To give another comparative example, AI is now used by CERN at the Large Hadron Collider to search, filter and categorise ~45 petabytes of data per week[15], the equivalent of 22.5 to 45 BILLION e-books, 45,000 –90,000 times the 500,000 e-book volume that the government says can now be found in a single criminal investigation.
Faster Justice, But at What Cost?
As with most emerging technologies, there is no shortage of justifiable scepticism when it comes to integrating AI with criminal justice. Those issues deserve a great deal more than cursory treatment. To keep the present discussion focused on whether AI has the potential to relieve pressures on court capacity, they are not examined fully here, but will be considered separately in subsequent articles. However, for completeness, I will summarise the principal concerns below.
- Accuracy and omissions
- Most would agree, when it comes to matters of criminal justice, ‘mostly right’ is not good enough. Missed evidence, incorrect identifications and erroneous assessments of case readiness go far beyond mere inconvenience. Such errors risk wasting court time, draining limited resources, and, in the worst case, risk innocent people being deprived of their liberty. Are these challenges already facing our justice system? Yes. But when it is AI that makes the mistake, who is held to account?
- The verification paradox
- Do we really want a system where AI can autonomously execute the criminal process? Probably not. But if the efficient work of AI tools needs to be checked and approved by humans, how much time is saved? We don’t know. Certainly less; perhaps not much at all. But remove human safeguards and the risk of errors inevitably increases. Such is the paradox.
- Equality of arms
- Particularly relevant to police and prosecution tools, such as PoliceAI, discussed earlier. Does equality of access to evidence mean much if legally aided defence practitioners are contending with publicly funded AI tools that can analyse, search, and sort the same evidence in seconds, as opposed to hours, days or weeks? Access may be the same, but practically their ability to process and use that evidence is worlds apart. This ultimately raises questions of the rights afforded to the parties under Article 6 of the ECHR.
- Automation bias
- This is a global issue affecting most industries and individuals. Humans are quick to believe that AI has it right. It is quite feasible that a highly experienced listing officer could take one look at a particular estimate of trial duration and feel sure it is inaccurate, but nevertheless he/she stays quiet, because tech is always right, right? Once we begin to assume AI is reliable, our own decision making is eroded and criminal justice become entirely AI led.
Conclusion
I am of the view that we CAN use AI to help solve the backlog of cases in the criminal courts of England and Wales, but it cannot do so on it’s own. It will no doubt take significant planning, building on the strong research and suggestions of Sir Brian Leveson and other practitioners and academics, substantial funding (AI tools and tech integration are not cheap), and a high level of technical training for the HMCTS staff that we expect to use these new systems.
Along the way, there will be many issues to be address to ensure that the adoption of AI does not undermine the integrity of the criminal justice system. To be clear, I believe that AI can augment the criminal process, relieving the administrative burden and better predicting when and where the courts limited capacity will be required. I do not, however, support its use as a substitute for the careful and considered judgement of judges and advocates. Ultimately, I believe that AI affords us a significant opportunity to reduce the bottlenecks seen throughout the criminal justice system whilst also preserving procedural fairness.
Those issues, of procedural fairness and the uptake of AI, will be the focus of a number of articles in the future.
[1] André Vasconcelos Roque, ‘A Luta Contra o Tempo Nos Processos Judiciais: Um Problema Ainda à
Busca de Uma Solução’ (2011) 7 Revista Eletrônica de Direito Processual 237, http://www.e-publicacoes .uerj
.br/index.php/redp/article/view/21125
[2] Ministry of Justice, Criminal Court Statistics Quarterly: January to March 2026 (25 June 2026), sections 1 & 2. https://www.gov.uk/government/statistics/criminal-court-statistics-quarterly-january-to-march-2026/criminal-court-statistics-quarterly-january-to-march-2026
[3] ibid
[4] ‘efficiency, n.’ (Cambridge Dictionary Online) https://dictionary.cambridge.org/dictionary/english/
efficiency, accessed 9 August 2026.
[5] Luisa Hedler, ‘AI and the Efficiency of Courts’ in Monika Zalnieriute and Agne Limante (eds), The Cambridge Handbook of AI and Technologies in Courts (Cambridge University Press 2026) 309.
[6] Ministry of Justice, AI Action Plan for Justice (31 July 2025) https://www.gov.uk/government/publications/ai-action-plan-for-justice/ai-action-plan-for-justice accessed 9 August 2026.
[7] ibid
[8] Sir Brian Leveson, Independent Review of the Criminal Courts: Part II, vol 1 (Ministry of Justice 2026) 68–69 [23]–[26], 190 [152].
[9] Ministry of Justice, HM Courts & Tribunals Service and HM Prison and Probation Service, ‘AI tech ambition to deliver smarter justice for victims’ (GOV.UK, 9 June 2026) https://www.gov.uk/government/news/ai-tech-ambition-to-deliver-smarter-justice-for-victims accessed 9 August 2026.
[10]Sir Brian Leveson, Independent Review of the Criminal Courts: Part II, vol 2 (Ministry of Justice 2026) 382–383 [80]–[83], Recommendation 95.
[11] Rémi Lam and others, ‘Learning Skillful Medium-Range Global Weather Forecasting’ (2023) 382 Science 1416, 1416–1421 https://www.science.org/doi/10.1126/science.adi2336 accessed 10 August 2026.
[12] Dan Benn, ‘AI to transform police evidence handling in landmark reform’ (Public Sector Executive, 15 July 2026) https://www.publicsectorexecutive.com/articles/ai-transform-police-evidence-handling-landmark-reform accessed 9 August 2026.
[13] Home Office, ‘AI to speed up justice under major disclosure reforms’ (GOV.UK, 14 July 2026) https://www.gov.uk/government/news/ai-to-speed-up-justice-under-major-disclosure-reforms accessed 9 August 2026.
[14] Ministry of Justice, Criminal Court Statistics Quarterly: January to March 2026 (25 June 2026), sections 1 & 2. https://www.gov.uk/government/statistics/criminal-court-statistics-quarterly-january-to-march-2026/criminal-court-statistics-quarterly-january-to-march-2026
[15] Antonella Del Rosso, ‘A new data centre at CERN’ (CERN, 23 February 2024) https://home.cern/new-data-centre-cern/ accessed 10 August 2026.
Leave a Reply