CLA News / A Case for Rooted AI Regulation in Commonwealth Courts By Nitish Rai Parwani

Introduction

Artificial intelligence tools are increasingly being deployed in legal systems around the world, and commonwealth countries are no exception. These AI tools are rapidly becoming structural components of legal systems across the Commonwealth. In the Caribbean, judiciaries are leveraging platforms like JUDI and AIDA[1] to automate document creation, streamline legal research, and manage case flows. In Africa, the Supreme Court of Kenya has spearheaded the development of a dedicated Judiciary Artificial Intelligence Adoption Policy Framework[2] to guide algorithmic integration while safeguarding judicial independence. In the Americas, Canadian administrative tribunals continue to experiment with automated triaging and predictive tools for case flow management.[3] Many of these deployments began largely in the absence of comprehensive regulatory frameworks, which are now being developed to regulate the technology.

In June 2026, the Supreme Court of India released a draft Regulations for Use of Artificial Intelligence in Courts, 2026 [4](hereinafter, ‘Indian draft regulations’, for this essay) for public consultation. India is one of the largest and perhaps among the most internally plural and diverse jurisdictions in the Commonwealth.  Against this background, this essay uses the Indian draft regulations as a case study to examine whether AI regulations in Global South Commonwealth jurisdictions—particularly those governing the deployment of AI in legal systems—reflect regulatory imaginaries rooted in their own intellectual and social realities. It further asks what approach may enable these countries to retain their uniqueness and plurality while developing globally credible standards for the regulation of AI in their courts.

The essay also considers whether and how such “universal principles” can be translated into the categories, ethical frameworks, and social realities of Global South societies and thereby become what this essay terms “rooted regulation”.

The legal systems of many Commonwealth countries have inherited substantial elements of normative frameworks developed “elsewhere”. Colonial-era laws often incorporated European legal and ethical assumptions without meaningful local participation in their formulation. Given limited political agency at the time, there was no other choice as well[5]. However, even after independence—including in the recent ‘reforms’ claimed as decolonization attempts—many of those foundations remained intact. AI governance presents a rare opportunity to depart from this trajectory—and initiate ‘rooted regulations—particularly when a large set of ‘global AI ethics norms’ that are framed and argued as ‘universal’ fail to recognize epistemic diversity[6] and to establish dialogical relations with non-Western knowledge systems, thereby reproducing colonial asymmetries of thought and governance[7] within technological design and deployment.

The case of the Indian draft

The Indian regulation draft, for instance, contains several safeguards, like prohibition on algorithmic adjudication and sentencing, bar on AI-based risk scoring in matters such as bail and recidivism, that deserve recognition. However, the draft largely adopts the jargon and vocabulary that have become globally dominant in AI governance, without reflecting on whether, and in what form, these concepts are appropriate to the Indian context or whether alternative conceptual foundations were considered. Apart from a cursory reference to the Constitution, there is little attempt to derive AI governance principles from specifically Indian intellectual resources. There is no engagement with indigenous or Dharma-based jurisprudential traditions, little discussion of pluralism beyond accessibility, and no explicit recognition of India’s social heterogeneity as a governance challenge.

India’s social reality is characterized by the coexistence of multiple linguistic, cultural, religious, and philosophical traditions; a reason also why it serves as an ideal case study for the global south and commonwealth countries with different cultural practices and pluralistic social fabric. A governance framework designed for such a society should be attentive not merely to general ‘universal’ principles, but also to plurality, i.e. not simply identifying abstract norms that apply everywhere, but to develop institutions capable of accommodating diverse ways of understanding justice, dignity, and social order[8]. The universal principles, if any, must be translated to local contexts. This is particularly important for courts deciding domestic cases, because disputes arise within local contexts and remedies must ultimately be framed with reference to those contexts.

Dharma based framework for sustaining plurality

Dharma-based legal and governance traditions provide historical examples of frameworks through which plurality could be accommodated within broader normative structures across parts of South, South-East and East Asia[9]. A common misunderstanding of Dharma made a categorical error of translating it as ‘Religion’. In practice, Dharma has been a much wider connotation, encapsulating ethics, regulations, law, policy logic, conduct and order. [10]It is not necessarily theistic, and accommodates diverging views of philosophy. In their operations, Dharma based frameworks acknowledged localized layers of authority and ethical standards (deśa-kāla-paristhiti) within broader universal ethics. They allowed diverse sections of society to maintain and practice their respective worldviews (and achieve justice according to them) and customs, sans an attempt on homogenization. The colonial experience substantially displaced or flattened these structures within formal state institutions, although elements of them survived in other social domains and contributed to the retention of distinctive cultural identities.  Recent initiatives—including the works and advisory of Oxford Dharma Centre’s AI and Dharma Ethics Programme[11]—illustrate growing interest in exploring these ideas.

A Case for multi-agent frameworks

The current regulations on AI in courts, including the Indian draft, occasionally recognize diversity, use the terms like Fairness, non-discrimination, which are now also the part of ‘universally accepted’ standards to contain bias in the AI systems. However, they may be technically and philosophically difficult to implement. The AI systems’ algorithms are not built in a cultural or political vacuum; and may be designed and/or trained on datasets reflecting historical and prevalent biases.[12] This may be pronounced more in the formerly colonized countries, where the bias in governance—stemming from imperial over indigenous interests, widening of fault lines of cultures, etc.—would be recorded in datasets.[13] This may also include the governance principles inheriting legacies of colonial institutions. Thus, courts relying on such systems may not be able to aptly capture the aspirations and contexts of the subjects and the State of their operation.

While the opportunity to actively create a new system integration with judicial systems is available, there may be provisions for a multi-agent framework, where, instead of presenting one “fair” answer, AI systems could expose multiple and even competing lines of authority, alternative characterizations or possible narratives of the relevant facts, or different legally plausible interpretations, and allow the human decision-maker to exercise judgment. This may prevent ‘anchoring effect’ in decision-making[14] by avoiding just one perspective which may anchor a decision-maker’s cognition, and mitigate automation bias, whereby a user may become overly reliant on AI-generated suggestions.[15].

Alignment with local traditions and epistemic representation

Many regulations, including the Indian draft, acknowledge inclusivity and non-discrimination; but do not necessarily provide any operational mechanisms for addressing the deep diversity. Emerging Global-South scholarship, particularly in Indian academia, has been exploring identity-sensitive and context-dependent approaches to AI value alignment, suggesting that ethical reasoning cannot always be reduced to a single universal optimization framework.[16] At a time when discussions of indigenous knowledge systems, epistemic diversity, and intellectual decolonization occupy a prominent place in academic and policy discourse, one might have expected a framework of this significance to engage more seriously with scholarship, and the possibility that India could contribute its own conceptual resources to the governance of AI.

Regulations being developed in Global South Commonwealth jurisdictions should not simply make room for the borrowing of datasets and algorithmic logics developed elsewhere. They should establish substantive standards for determining what constitutes an adequately representative dataset for their own jurisdictions and how such representativeness should be assessed. The audit of the datasets should be periodical and independent, to check perpetuation of any bias. The training datasets must be assessed for “Epistemic Representativeness’—equitable accommodation of plurality they represent, and thus formally acknowledge that the legal system operates across multiple cultural, linguistic, and historical layers which exist within these jurisdictions.

Several regulations also omit concerns regarding collective and community-level harms. Borrowing from ‘international standardizations’ of privacy and personal data protection, they limit to preventing the identification of individuals.[17] Algorithmic harms, however, often operate at the level of groups rather than individuals. AI systems can disproportionately affect communities, linguistic groups, or socio-economic categories even when no specific individual is identifiable. Hence, given diversity in these aspects, the commonwealth jurisdiction must be conscious of collective harms.

Experiences from predictive policing and algorithmic governance demonstrate how systems trained on historically skewed datasets can reinforce patterns of exclusion, surveillance, or disadvantage at the group level while remaining formally compliant with individual privacy protections. Protecting individual identities does not necessarily prevent the stigmatization of communities. A governance framework suited to India’s (and Global South Commonwealth jurisdictions’) social realities should therefore be attentive not only to individual harms but also to the systemic and communal consequences of algorithmic decision-making. A “Community Impact Assessment” may be conducted before any tool is deployed, ensuring that predictive or administrative AI does not disproportionately flag, ignore, or marginalize specific socio-economic groups. Such an assessment could form part of the pre-deployment approval process and examine whether a system’s training data, classifications or outputs disproportionately exclude, misrepresent or burden linguistic, regional or socio-economic groups. Where material disparities are identified, deployment of such a model or training data should be withheld or made conditional upon mitigation and subsequent independent review.

Conclusion

As the Artificial Intelligence is finding roots in the justice delivery systems, it is important to regulate its scope and contours. For this purpose, principles presented as “universal”, but developed in particular foreign contexts, may provide a persuasive regulatory baseline; they cannot, however, by themselves constitute a complete framework for judicial AI governance.  The attempts to self-draft regulations on the use of AI in Global-South commonwealth jurisdictions represent an important beginning. Now it is to decide whether AI governance, just like legal and governance frameworks of the past, will continue to borrow its normative assumptions from ‘elsewhere’ or whether it will draw upon its own intellectual resources and unique experiences to articulate a distinctive and globally relevant model of technological governance. While these jurisdictions should remain open to international standards and comparative experience, there must be rigorous translating and assimilation of guidelines through their own social realities, intellectual traditions, linguistic and normative diversities, constitutional structures, and institutional capacities. Such a framework, a rooted framework, would ensure that borrowed principles acquire legitimacy and practical meaning within the order they are intended to govern. This is more relevant in the context of courts of law, as they are arbiter—often final arbiter—on formal actions in a state.

Nitish is a D.Phil scholar at the University of Oxford. He completed LL.M from National Law University Delhi, and clerked with a former Chief Justice of India. He also coordinates the Oxford Dharma Centre for Research and Policy.

[1] AI in the Courts: Caribbean Innovations in Action available on: https://connectedcaribbean.org/ai-in-the-courts-caribbean-innovations-in-action/

[2] Judiciary to leverage AI to enhance justice available on : https://judiciary.go.ke/judiciary-to-leverage-ai-to-enhance-justice/

[3] Eg: https://www.lexisnexis.com/blogs/en-ca/b/canadian-product-spotlight/posts/artificial-intelligence-and-administrative-law-in-canada-striking-the-right-balance.

[4] Seeking views/suggestions of all stakeholders and the general public on draft ‘Regulations for Use of Artificial Intelligence (AI) in Courts, 2026 available on : https://cdnbbsr.s3waas.gov.in/s3ec0490f1f4972d133619a60c30f3559e/uploads/2026/06/2026060342.pdf

[5] The legislative function was mostly taken over by the colonial governments and agencies, with limited or no agency with the local voices. In India, for instance, the codifications and legislations undertaken during colonial period—including by the British Parliament in London—continued even after political independence in 1947.

[6] A Lens of Dharma for the Emergent AI-Ethics Discourse: A working paper

available on : https://www.oxforddharma.org/dharma-lens-ai-ethics-discourse.pdf

[7] AI ethics through a decolonial lens: what does AI ethics look like if we take seriously the push to decolonise it? Available on : https://link.springer.com/article/10.1007/s00146-026-02935-9

[8] Pluralism in AI Governance: Toward Sociotechnical Alignment and Normative Coherence

Available on : https://arxiv.org/abs/2602.15881?

[9] Eg: Voyce, M. (2007). The Vinaya and The Dharmaśāstra: Monastic Law and Legal Pluralism in Ancient India. The Journal of Legal Pluralism and Unofficial Law, 39(56), 33–65; Menski, W. F. (2010). Sanskrit Law: Excavating Vedic Legal Pluralism. SOAS School of Law. Research Paper, 05-2010, 1-44.

[10] Parwani, N.R. (2025), Dharma: Sustainability principle of all relationships, Prabuddha Bharata, Jan. p. 157-162.

[11] AI & Dharma Ethics Programme (AiDEP) available on : https://www.oxforddharma.org/programmes/aidep

[12] Hanna, M. G., Pantanowitz, L., Jackson, B., Palmer, O., Visweswaran, S., Pantanowitz, J., Deebajah, M., & Rashidi, H. H. (2025). Ethical and bias considerations in artificial intelligence/machine learning. Modern Pathology, 38(3), Article 100686.

[13] Muldoon, J., & Wu, B. A. (2023). Artificial intelligence in the colonial matrix of power. Philosophy & Technology, 36, Article 80.

[14] Nguyen, J. K. (2024). Human bias in AI models? Anchoring effects and mitigation strategies in large language models. Journal of Behavioral and Experimental Finance, 43, Article 100971.

[15] Romeo, G., Conti, D. Exploring automation bias in human–AI collaboration: a review and implications for explainable AI. AI & Soc 41, 259–278 (2026).

[16] Eg: Madva, A., Srivatsa, S., Srinivasa, S., & Saha, T. (2026). Rethinking Indic AI from a lens of cultural heritage preservation (arXiv:2607.06544v1). arXiv. https://doi.org/10.48550/arXiv.2607.06544;

[17] Graef, I., & van der Sloot, B. (2022). Collective Data Harms at the Crossroads of Data Protection and Competition Law: Moving Beyond Individual Empowerment. European Business Law Review, 33, 513–536.