AI in Courts Could Open Doors to Justice or Leave Vulnerable People Further Behind

UNESCO’s account, published on 28 September 2026, highlighted the need to judge new technology by its effect on people’s access to justice.

For a farmer unable to afford a journey to court or a domestic violence survivor living far from legal services, a digital justice system is only useful if they can actually reach it. That concern shaped a panel at the fourth UNESCO Global Forum on the Ethics of Artificial Intelligence in Riyadh, Saudi Arabia, where speakers examined how AI could make justice more accessible and the decisions that could push vulnerable people further away. UNESCO's account, published on 28 September 2026, highlighted the need to judge new technology by its effect on people's access to justice.

Digital Courts Need to Reach People Beyond the Screen

Alpha Sesay, Sierra Leone's Attorney General and Minister of Justice, described a country where many people resolve their justice needs outside formal courts and less than 30% of the population has mobile internet access. Moving services online without considering those realities could create new barriers for people already struggling with distance, cost or personal safety. Sesay called for digitalisation to begin with identifying who might be excluded, including rural domestic violence victims and farmers who cannot afford to travel.

Judge Abdulaziz bin Mohammed Al-Samaani of Saudi Arabia's Board of Grievances said judicial AI strategies should improve access to knowledge and services without weakening the fundamental safeguards of justice. Rwanda's Chief Digital Officer at the Ministry of Justice, Stella Murungi, described a practical way to extend digital access through IremboGov's community agents, who operate from local shops and help rural residents use government services at harmonised prices. She highlighted women's representation in decision-making, citing Rwanda's constitutional minimum of 30% and a share of around 50% among judges.

Everyday Court Decisions Can Carry Hidden Costs

A UNESCO survey found that 44% of judicial operators worldwide were already using some form of AI, although only 9% had received formal training, exposing a substantial gap between adoption and preparation. Smita Mutt, Strategic Initiatives Lead at India's civil society and research organisation DAKSH, warned that exclusion becomes more likely when technology spreads faster than the rules and oversight needed to govern it. Her concern extended to routine court administration, where decisions about timing can affect whether people can participate effectively.

Court lists published late in the evening can place a particular burden on female advocates with caregiving responsibilities, and last-minute rescheduling can leave vulnerable litigants exposed to further difficulties. Mutt argued that assessments need to examine these less obvious consequences, giving attention to the people who bear the costs of administrative decisions. A scheduling system may appear efficient from inside a courthouse, even as unpredictable arrangements make attendance and preparation harder for those managing work, care responsibilities or personal risks.

Human Oversight Must Remain Central to Judicial AI

Judge Abdulaziz stressed that AI-generated outputs need verification, a concern illustrated by an early case in Colombia in which a judge acknowledged using a generative AI tool's reasoning in a decision about a child's insurance benefits. The example raised questions about the consequences an inaccurate or exclusionary response could have for someone seeking justice. Murungi said Rwanda checks AI-generated language before deployment, including translation accuracy for non-English speakers, and distinguishes between low-risk administrative tasks suitable for automation and high-risk uses such as sentencing, where a human remains the final decision-maker.

Speakers called for digital services, community participation and governance to work together, with safeguards established before deployment. UNESCO presented its Ethical Impact Assessment methodology and AI Readiness Assessment Methodology as practical resources, alongside its free course on AI, Justice and the Rule of Law, developed with the University of Oxford. The panel's central test for success was whether people most likely to be excluded can obtain the justice they need, with investment in technology matched by attention to access, training, reliable information and human responsibility.