2026 YLAL Student Essay Competition, and the winner is… Gvantsa Gvadzabia!
The scores for the 2026 YLAL Student Essay competition are in, and the winner is… Gvantsa Gvadzabia! (see profile linked).
You can read Gvantsa’s insightful essay on the impact of AI on access to justice in legal aid services below:
Critically analyse the ways in which artificial intelligence might increase, undermine, or fundamentally transform access to justice for those who rely on legal aid services in England and Wales.
AI complicates what equality of arms requires in a legal system marked by unequal resources. Greater access to legal information may reduce one barrier without improving a person’s practical capacity to vindicate their rights against a better-resourced opponent. The relevant measure is not how much legal assistance AI can produce, but whether it narrows that imbalance. Presently, it is more likely to widen it because technological capacity is itself unequally distributed.
AI may transform an old access-to-justice problem – the imbalance of power between the individual and the state. If legal aid exists to prevent economic inequality from determining whether state power can be challenged, technological inequality should not be permitted to achieve indirectly what financial inequality is supposed to be prevented from achieving directly. By 2025, Public Law Project had identified 55 algorithmic tools used in administrative decision-making, demonstrating the existing scale of automated government. Scale becomes an access-to-justice problem when the power to automate decisions is not matched by the capacity to interrogate them. Research submitted to Parliament found legal-aid organisations struggling to assist benefit claimants flagged for fraud investigation by DWP data-matching algorithms because insufficient information was provided about why they had been selected. This exposes a problem deeper than accuracy – contestability. Even a technically sound system may impede justice if those affected cannot understand or effectively challenge its application to them. That deficit is especially significant within a legal-aid system from which LASPO removed much early social-welfare advice; the Justice Committee has since found that reduced opportunities for early intervention allow problems to escalate before assistance becomes available. AI therefore risks increasing the complexity of administrative power precisely where capacity for early legal challenge has contracted. However. scalability need not favour the state. Technology could enable limited legal-aid practitioners to scrutinise greater volumes of decisions and expose patterns invisible case-by-case, potentially scaling accountability too. The difficulty is distributive – the Justice Committee now finds that financially strained providers lack capacity to invest in new technology. AI could therefore strengthen legal-aid lawyers’ capacity to challenge the state while simultaneously making the state they challenge more technologically complex. Whether it increases access to justice depends on which of those capacities develops faster.
The strongest objection is that this diagnosis may invert AI’s significance – the same scalability that strengthens institutional power could allow scarce legal expertise to reach those whom the existing system excludes. Around 32% of adults experiencing a legal problem receive no professional support, often because of cost, complexity or uncertainty about where to turn. For them, the relevant comparison may not be AI against a publicly funded lawyer, but AI against no assistance at all. This potential extends beyond information provision. The Legal Aid Agency is already piloting AI to navigate complex LASPO legislation and guidance, reporting approximately 95% accuracy within weeks. AI could therefore scale not merely legal information, but the capacity of limited practitioners to provide it.
Yet equal access to computation is not necessarily equal access to legal capability. AI’s usefulness and accountability also depend upon the information against which its outputs can be tested. First-tier social-security adjudication, for example, lacks an equivalent comprehensive public database of decisions, while DWP, as a repeat institutional party, has received tribunal feedback explaining why its decisions were overturned and used it to improve decision-making. This does not establish that AI is less accurate in social-welfare law; legislation and appellate authority remain available, while first-instance tribunal decisions are not binding precedent. Instead, it exposes the limit of the democratisation claim – giving both sides access to similar computation cannot equalise institutional data, accumulated feedback, or professional expertise. AI may therefore be genuinely transformative where the principal barrier is obtaining or processing accessible legal information; its equalising force diminishes where effective challenge depends upon informational resources that technology itself does not distribute.
The democratisation argument therefore succeeds where scarcity itself is the principal barrier to justice – if AI enables an overstretched adviser to assist more clients or provides guidance where none was realistically available, access has increased. But its logic stops where information is not the decisive inequality. A person may understand the law yet still lack the evidence, institutional knowledge or professional capacity required to contest an opponent’s decision. Nor is the asymmetry inevitable – if legal-aid practitioners use AI to interrogate automated decisions, identify systemic patterns or remove routine work, technology can narrow rather than widen the imbalance. The question is therefore not whether AI expands access to law, but whether it redistributes the capacity to use law against power.
On that measure, the negative case presently carries greater weight. Legal aid exists precisely because information alone cannot overcome inequalities in resources and representation. Yet the institutions exercising power are presently better positioned to capture AI’s advantages than the publicly funded infrastructure challenging them: financially strained providers lack investment capacity, while consumer-facing AI protections remain weakest around vulnerability, referral to appropriate assistance and redress. The risk is therefore distributive, not technological. AI may augment professional judgment for those able to obtain it while becoming a substitute where publicly funded assistance is unavailable. Rejecting AI would only entrench this disparity. Until legal aid can scale accountability as effectively as institutions scale power, AI is more likely to widen than close the justice gap.
AI is therefore presently more likely to widen than close the justice gap. Yet its strongest case in legal aid arises from the dangers AI itself creates: as state power becomes technologically augmented, equality of arms increasingly requires technological capacity to challenge it. AI may thus transform access to justice not by replacing legal aid, but by becoming necessary to its effectiveness.