Must not replace human judgement
As criminal courts buckle under sustained pressure, an independent review has urged the adoption of AI to improve efficiency and reduce mounting backlogs.
The second part of Sir Brian Leveson’s Independent Review of the Criminal Courts, the first of which was published last July, focuses on improving the efficiency of the courts system in order to tackle the backlog which has been plaguing the courts system for years.
In it, Leveson, former president of the Queen’s Bench Division, sets out a wide-ranging package of recommendations aimed at restoring efficiency to a system where justice is increasingly being “delayed and thus denied”.
Most noticeably, as part of his whopping 135 recommendations, Leveson affords significant attention to the integration of AI into the criminal justice system, ranging from general automation and productivity tools to its use in translation for defendants who can’t speak English.
In the report, Leveson goes to great lengths to emphasise the appropriate use cases of AI and warns against what he calls “predictive: tools, namely using AI to analyse data in order to forecast future outcomes. According to Leveson, such systems carry too much potential for bias and should not be used in a way that replaces human judgment. He also discusses the broader risks AI poses due to the limitations of the technology, such as its tendency to hallucinate and its potential for bias.
Of the recommended uses for AI, Leveson instead argues that it should be deployed “solely to augment rather than replace human decision-making”, with acceptable applications falling into three broad categories: productivity tools, insight tools, and accessibility tools.
Firstly, AI could be used to enhance efficiency by automating daily administrative tasks, such as drafting documents and summarising case files, through the integration of tools such as Microsoft Co-pilot “to support staff in managing their day-to-day responsibilities.”
Leveson also outlines specific recommendations for each step of the criminal justice process.
For instance, under recommendation 80, Leveson outlines the potential for the use of AI in the Crown Prosecution Service (CPS), helping prosecutors analyse evidence and material to aid in decision-making, with the potential ability to spot gaps in the case file such as missing documentation.
Leveson also discusses the potential use cases of AI within disclosure, particularly in relation to processing and analysing large volumes of material.
He also highlights the potential of AI-enabled translation tools in hearings, as well as the potential of AI-enabled transcription tools, which could create accurate transcripts more swiftly and at lower cost.
Leveson additionally recommends the use of AI to summarise the Initial Details of the Prosecution Case, arguing that this could “facilitate His Majesty’s Courts and Tribunals Service staff and other relevant partners in their preliminary reviewing roles.”
Given the broad scope of the review, Leveson also offers recommendations for the use of artificial intelligence within the wider criminal justice system, such as in the process of drafting witness statements by the police.
These recommendations come at a time when concerns are increasingly being raised about AI’s role in the legal sector, with the outgoing President of the Supreme Court, Lord Reed, warning that AI could threaten trust in the court system, alongside growing fears about the impact of hallucinations and “slop” on the profession.