Human Primacy in the Age of Algorithms – The Supreme Court’s Draft AI Regulations for Courts, 2026

The interface between emerging technology and institutional justice delivery is an increasingly important area of Indian legal policy. By a notice dated 3rd June 2026, the Supreme Court of India, through its Artificial Intelligence Committee, released for public consultation a draft “Regulations for Use of Artificial Intelligence (AI) in Courts, 2026,” inviting comments up to 20th June 2026.

Grounded in human primacy, transparency, accountability, data protection, and judicial independence, the draft governs AI deployment across the Supreme Court, High Courts, and all Courts, Tribunals, and Commissions performing adjudicatory functions. It spans ten chapters and fifty-seven regulations, resting on five foundational principles:

• Human primacy
• Transparency
• Accountability
• Data protection
• Judicial independence

Background

The draft follows deliberation by the AI Committee on the growing use of AI in transcription, translation, research, and case management, and the resulting need for a governance framework.

Once notified, it will apply to the Supreme Court on a date fixed by the Chief Justice of India, and to each High Court’s jurisdiction on a date fixed by its own Chief Justice. Different provisions are permitted different commencement dates.

Competing Considerations

The draft reconciles two pulls. Regulation 16 creates a “presumption in favour of responsible AI adoption,” and Regulation 17 (“Innovation over Restraint”) directs that AI development furthering judicial efficiency be actively encouraged.

Conversely, Regulation 4 declares AI “strictly subservient to human judgment,” barring any system from supplanting judicial authority. The draft resolves this tension through calibrated safeguards rather than blanket rules in either direction.

Regulatory Scheme

General Principles and Human-in-the-Loop

Chapter II’s general principles anchor the balance between innovation and judicial authority. The most notable is the Human-in-the-Loop requirement (Reg. 3(zb)), mandating human review of every AI output, with accountability resting exclusively with the human officer.

Regulation 8 bars reliance on “Black Box” opacity or “hallucination” as a defence to a wrong decision. Officers may skip verification for recorded reasons, and certified administrative tools are deemed compliant class-wise.

Permissible Uses

Chapter III separates permissible uses from prohibitions. The following uses are allowed, subject to prior approval and supervision:

• Case management
• Transcription
• Translation
• Legal research
• Accessibility tools

Absolute Prohibitions

Regulation 20 sets out prohibitions that are expressly “absolute and non-derogable”:

• No judicial outcome may be reached by Algorithmic Decision-Making alone.
• No AI may adjudicate or sentence without human oversight.
• Risk Scoring (flight risk, recidivism, bail, credibility) is barred entirely.
• AI may not profile future conduct.
• AI-based surveillance of judicial officers, advocates, or litigants is prohibited.

Institutional Structure

A tiered governance structure supports the regulatory scheme:

• An Apex Body
• Court-level AI Committees
• District-Judge-headed AI Secretariats
• A Centre of Research and Excellence (CoRE-AI)

This structure supports mandatory pre-approval Impact Assessments, annual in-house audits, an AI Incident Database, and fall-back protocols.

Key Mandates

The draft imposes several cross-cutting mandates across courts and stakeholders:

• AI output remains advisory only.
• Sensitive judicial data cannot go to external systems without written authorisation and must be anonymised before training use.
• Disclosure is mandatory for both Court-initiated AI use and AI-assisted litigant filings.
• Private vendors need prior approval and audit/liability safeguards, and cannot claim IP over tools built on judicial data.
• Litigants harmed by prohibited AI use may seek redress under Regulation 52.

Concluding Commentary

Even at the consultative stage, the draft signals a structurally cautious approach—enabling modernisation of an overburdened court system while keeping adjudication firmly human.

By making certain prohibitions immune even from the Chief Justice’s general relaxation power, and pairing explainability and in-house audits with a statutory grievance mechanism, the Committee has sought to future-proof judicial authority against erosion.

If finalised in this form, the framework may become a reference point for how far AI can enter adjudication without displacing the judge.