Mayank Suri & Dolly Bachani
The Supreme Court (SC) of India’s recent draft ‘Regulations for Use of Artificial Intelligence (AI) in Courts, 2026’ is a landmark step at acknowledging technological prowess of AI in the current day & age. Given that the SC is the highest judicial authority in the most populated country of the world, this act shows the potential pervasiveness of AI in judicial work.
It also raises critical questions about the very nature of adjudication, accountability, & the realistic limitations of human oversight. Our observations showcase that the Bangalore Principles of Judicial Conduct (adopted in 2002) mentioned within the draft regulations sufficiently covered the use of AI by judicial officers. To this extent, we come to the conclusion that the Regulations, therefore, only represent utility in governing the conduct of non-judicial officers of the courts.
Juxtaposing AI’s Assistive Role with the Principle of Rule of Law
The SC’s draft regulations assert on AI’s role in the courtrooms to be strictly assistive. This is done through establishing ‘human primacy and judicial independence’ as the first governing principle in Regulation 4. However, the wording used in Regulation 2 gives the impression that this principle may almost always be ignored given the serious temporal limitations existing on the time of those who work in courts.
Regulation 5 is titled ‘rule of law’ and prescribes that the use of AI in courts shall be:
Consistent with the constitution
Consistent with all other laws
Consistent with principles of natural justice
Not be inconsistent with due process, the right to a fair trial, equality before law, or access to justice.
Sub-regulation 2 of Regulation 5 then restates the binding nature of the Bangalore Principles on judicial officers. These principles are:
Independence
Impartiality
Integrity
Propriety
Equality
Competence and Diligence
In Lon Fuller’s The Morality of Law, the Rule of Law is envisioned as an amalgamation of formal institutions of society with the “inner morality of law” through the use of eight procedural principles: 1) there must be rules, 2) they must be prospective, not retrospective, 3) they must be published, 4) they must be intelligible, 5) they must not be contradictory, 6) compliance with the rules must be possible, 7) the rules must not be constantly changing, and 8) there must be congruence between the rules as declared and as applied by officials (Fuller, 1969, p.39). Since the judiciary and the recipient sit in relation to each other, a justified balance is necessary between the Judiciary’s application of the Rule of Law and its effect and perception on recipients. This is where the Regulations read with the Bangalore Principles must meet Fuller.
While the Regulations mention that the Rule of Law ought to be followed, it is another enquiry on whether it can be followed or if it can even be perceived to be followed. Can we say for sure that an over-reliance on AI-generated case summaries or automated translations (both are permissible uses under the Regulations) won’t erode a judge’s ability to engage with the raw facts of a case? Let’s take Fuller’s latter 5 principles and transpose them upon the Regulations with the added applicability of the Bangalore Principles.
Fuller’s latter 5: intelligible, non-contraditory, possibility, static, and functional
When analysing Fuller’s 5 above, Regulation 7 is our first port of enquiry. This regulation mandates that the AI used in courts shall be transparent and explainable. It would seem that the very act of drafting this requirement into regulation indicates the mischief suffered by AI systems. This prima facie contradicts the intelligible nature requirement of the Rule of Law, something presumable in human interactions but yet to be sufficiently trusted in the case of AI. Its satisfaction would be dependent on the efficiency of the AI system in transparently explaining the decision taken. Transparency would require a reasonable amount of reasoning with procedure and precedence, and explainability would require it to be communicated in a form that is understood by the recipient.
Then Regulation 8 makes a human officer accountable for the decisions made with the use of AI for a palpably incorrect, illegal, or harmful decision. This stands to validate Fuller’s emphasis on the non-contradictory principle. Whether this principle will be satisfied will, arguably, be based on the number of correct decisions that are relied upon by such officers. The gambling nature of this adoption, however, is easy to notice. Whereas Fuller’s and the Bangalore Principles made certain values imperative, the Regulations make them unacceptable defences. It adds a layer. The difference will undoubtedly reduce competence and diligence.
If Fuller’s last 3: possibility, static, and functional are taken together they represent the epistemological grounds upon which recipients see the judiciary’s actions. Empathy, understanding, and realism, are imbued into the principles so that recipients achieve access to justice in their subjective circumstances. While non-judicial officers in person can be expected to make certain reasonable discounts to afford access to minorities and the unprivileged. Whether AI systems will follow depends on the effective integration of Regulation 13 on inclusivity and accessibility.
Conclusion
In our view, the draft Regulations are well intended and drafted. However, their effect is overshadowed by the well known principle of the Rule of Law. While this comment may not be treated as substantial, it is felt by us that an early adoption of AI in courts should be premised on a systemic reform of the judiciary at the human level. It is hoped that this is what the SC intended in incorporating capacity building, training, and best practices in Chapter VIII.