Nuffield Family Justice Observatory urges strengthened governance of AI use in courts to ensure litigants and system are protected
- Details
A report by the Nuffield Family Justice Observatory (NFJO) has warned that the Ministry of Justice (MoJ) and the judiciary must strengthen the governance of AI use in courts, to ensure that both litigants and the system are protected.
The report observes that the Government and court service are increasingly delegating digital and AI-mediated justice services to private providers, from digital legal information and dispute resolution tools to systems used within justice agencies.
“This shift is best understood as ‘delegation by default’ rather than design, driven by financial pressures, challenges with court reform and the rapid development of generative AI”.
In family justice, the NFJO warned that the risks are “particularly acute” because of the nature of disputes and the vulnerability of those involved.
The report notes that within courts, judges have been issued with a secure version of Microsoft Copilot.
However, there is “limited public information” about how its safe use was determined, which tasks it should be used for, how its use is monitored or how litigants can challenge its use.
The NFJO said: “General-purpose large language models (LLMs) used for summarisation and editing are prone to inaccuracy, hallucination and subtle shifts in meaning, including the flattening of ‘justified asymmetry’ that risks compounding courts’ documented tendency to reframe domestic abuse allegations as mutual conflict.
“Existing safeguards rely on judicial users of the technology as a ‘human in the loop’, but behavioural research shows this safeguard is unreliable in practice: the superficial fluency of AI outputs makes users less likely to scrutinise them even under incentives to check.”
The report also highlights evidence that litigants are turning directly to GenAI for advice and document preparation, despite these tools “being capable of getting the law wrong”.
The NFJO warned: “Their tendency to affirm users’ existing beliefs may be particularly problematic in emotionally charged family disputes, potentially entrenching conflict rather than supporting resolution. These tools may also enable vexatious litigants and domestic abusers to produce superficially convincing court documents at scale, placing further pressure on courts.”
In light of its findings, the NFJO called for the MoJ’s existing AI Action Plan to be “reconciled” with the HMCTS and Judicial AI Working Group’s principles and the Position Statement on the Judicial use of artificial intelligence, so that AI use is “governed by a coherent rather than competing framework”.
Meanwhile, it called for the MoJ, HMCTS and the judiciary to publish their evidence base for “disregarding Microsoft’s own advice against using Copilot where it could affect legal status or rights”.
The report makes a total of 26 recommendations across the following four themes:
- close regulatory gaps around general-purpose and domestic AI tools that provide legal advice or assistance, including creating a National Commission on the Regulation of AI in Justice;
- address market failure in consumer-facing tools by investing in trusted public legal information and developing quality standards;
- strengthen assurance and transparency around AI used within courts, including publishing evidence, approved tools and findings from completed pilots, and establishing independent oversight;
- strengthen governance of AI use in courts through coherent standards, monitoring and disclosure, safeguards that allow litigants to contest AI use, and ongoing evaluation of its impact on court capacity, case outcomes and public trust.
The NFJO concluded: “The report is not an argument against digital or AI-mediated justice, or against private-sector involvement. Instead, it argues that the current model of ‘delegation by default’ is a policy choice and should be replaced with a deliberate approach based on publicly governed standards for safety, performance and evidence, supported by effective oversight and accessible redress.
“Without action, government risks not only missing the potential benefits of AI and digital technology for access to justice, but creating avoidable costs, undermining public trust and compounding the problems digital justice is intended to address.”
Lottie Winson
Corporate Director of Legal and Regulatory Services
23-11-2026
08-12-2026 10:00 am

