Algorithmic bias in judicial processes poses a threat to constitutional guarantees of equality and liberty. In the context of India's draft court AI regulations, discuss the challenges of ensuring fair and explainable AI in the justice system.

Q. Algorithmic bias in judicial processes poses a threat to constitutional guarantees of equality and liberty. In the context of India's draft court AI regulations, discuss the challenges of ensuring fair and explainable AI in the justice system. (15 marks, 250-350 words)

The Supreme Court's draft Regulations for Use of Artificial Intelligence in Courts, 2026, released for public comment in June 2026, require AI to remain "strictly subservient to human judgment and judicial authority" [1]. This caution is well-founded: biased or opaque algorithms in adjudication can quietly erode Article 14 equality and Article 21 liberty.

How the draft guards against algorithmic bias - Bars AI from determining judicial outcomes, sentencing and bail decisions, and from predictive risk scoring — the very functions where bias translates into loss of liberty [1]. - Prohibits profiling of parties and witnesses, shielding women, minorities and marginalised litigants from algorithmic stereotyping [1]. - Bans "opaque or unexplainable" systems, protecting the right to a reasoned order under natural justice [1]. - Regulation 43 obliges lawyers and litigants to disclose AI-assisted filings, enabling scrutiny [1].

Challenges in ensuring fairness - Tainted training data: models learn from records in which roughly three-fourths of prisoners are undertrials, disproportionately from SC/ST/OBC and poor households [4] — the algorithm would encode this pattern as "risk". - Proxy discrimination: removing caste or religion does not help when address, occupation or prior arrests act as substitutes. - Even permitted assistive uses shape outcomes — research ranking and summarisation decide what a judge reads. - Profiling and case data raise data-protection duties under the DPDP Act, 2023 [3].

Challenges in ensuring explainability - Large language models remain black boxes; their tendency to fabricate citations has already drawn judicial censure. - Proprietary vendor models resist audit, clashing with disclosure norms. - Capacity gaps: uniform certification across High Courts and thousands of subordinate courts demands technical skills and infrastructure the reconstituted SC AI Committee and proposed Apex Body must still build [2].

The draft thus draws the right constitutional line — efficiency through AI, adjudication by humans. Its promise now rests on execution: independent bias audits, explainability standards for court-approved tools, judicial digital literacy, and early notification of the regulations. Done well, technology becomes an instrument of access to justice rather than a threat to equality before law.

(~330 words)

Sources: 1. Draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026 — Supreme Court of India (Notices & Circulars) — subservience principle, permitted/prohibited uses, ban on opaque AI, Regulation 43 disclosure 2. Supreme Court Reconstitutes AI Committee to Oversee Adoption and Deployment of AI Tools — News on AIR (Dec 2025) — committee-led oversight and implementation across courts 3. The Digital Personal Data Protection Act, 2023 — MeitY — data-protection obligations over personal data of parties and witnesses 4. Prison Statistics India — National Crime Records Bureau — share of undertrials and their socio-economic composition