When the Supreme Court’s Artificial Intelligence Committee put out a 30-page draft framework for regulating AI use across India’s courts, the headline provision was simple enough: any lawyer or litigant using AI to prepare a pleading, document, or piece of evidence would have to say so, in writing, at the time of filing. Three months later, that one requirement — not the parts of the draft dealing with judicial adjudication, data protection, or algorithmic audits — is the part practising lawyers, a sitting High Court judge, and the tech industry have all pushed back on hardest. The Draft Regulations for Use of Artificial Intelligence in Courts, 2026, released for public comment on June 3, are still unfinished business: the comment window has closed, but the rules have not been notified, and the fight over mandatory AI disclosure is a live preview of how India plans to govern AI inside its justice system.

What the draft actually regulates

The AI Committee’s framework is built around five stated principles — human primacy, transparency, accountability, data protection, and judicial independence — and it applies broadly: to the Supreme Court, every High Court, and “every court, tribunal, and statutory commission performing adjudicatory functions across the country,” per the committee’s own explanatory note. It does not switch on everywhere at once — implementation dates are to be notified separately by the Supreme Court and each High Court.

Under draft Regulation 19, AI tools are explicitly permitted for legal research, citation verification, drafting assistance, cause-list preparation and hearing scheduling, transcription of proceedings, translation of judgments, and case and record management — the administrative and research-support layer of court work. Regulation 20 draws the opposite boundary: AI cannot decide a case, determine bail eligibility, assess witness credibility, or be used to monitor judges, advocates, or litigants unless a separate law authorises it, and “opaque or non-explainable” systems are barred wherever legal rights or personal liberty are at stake. A permanent apex body — envisaged with two Supreme Court judges, two High Court Chief Justices, two more High Court judges, a Joint Secretary from the Ministry of Electronics and IT, and outside experts in finance, cybersecurity, and AI — is meant to set standards, approve which AI systems courts may actually use, and publish annual reports on how the framework is working in practice, per LiveLaw’s summary of the draft.

None of that framework structure has generated much public friction. What has is a narrower provision buried deeper in the text.

The disclosure clause at the center of the fight

Regulations 43(3) and 43(4) require an advocate to disclose “the AI-assisted character” of a pleading, document, or piece of evidence through a formal certificate whenever an AI tool was used to prepare it — not just for generative drafting, but potentially for research and citation-checking too. The Supreme Court initially gave the public until June 20 to comment, then extended that deadline to July 15 after stakeholders asked for more time. It is that extra month of submissions that produced the sharpest objections.

SCAORA’s objection

The Supreme Court Advocates-on-Record Association filed its comments on July 19, asking the Committee to drop the disclosure requirement outright rather than merely narrow it. SCAORA’s core argument is that the obligation is redundant: advocates, it said, “are officers of the court already bound by strict statutory codes of professional conduct” under the Advocates Act, 1961 and the Supreme Court Rules, 2013, and are personally responsible for the accuracy of every filing regardless of how it was prepared. Layering a separate AI-specific certificate on top, the association argued, is “unjustified” and risks becoming a box-ticking exercise rather than a real safeguard — while also carrying an unaddressed risk of prejudicing a litigant’s case or subtly biasing a judge who reads the disclosure.

Rather than leave a gap, SCAORA proposed a narrower substitute: amend the existing Special Leave Petition certificate to add one line, certifying that “all legal citations, judicial precedents, and statutory authorities relied upon in this Petition have been personally verified for accuracy” — shifting the obligation from disclosing how a document was made to confirming that it was checked. The association also flagged that several AI tools are already embedded in Indian court workflows — SUPACE, SUVAS, SuSahayak, and AI-assisted e-filing among them — and asked for phased audits of those existing systems, seats for practising advocates on the eventual governance bodies, and clearer definitions of undefined terms like “Court Data” and “High-Risk AI Tools.”

A judge’s public doubts

The disclosure clause has drawn skepticism from the bench as well. Speaking at the International ADR Conference 2026 alongside Senior Advocate J Sai Deepak, Delhi High Court Justice Prathiba M. Singh questioned whether the requirement was needed at all, warning that it would “increase [lawyers’] compliance burden and eventually become another set of templated affidavits” — paperwork filed reflexively rather than a meaningful check on anything, according to Bar & Bench’s report. Justice Singh’s broader point went beyond disclosure: she urged the judiciary to build its own dedicated AI platform rather than routing sensitive case material through public AI tools, warning that feeding matrimonial or criminal case details — often containing sensitive personal allegations — into general-purpose AI products could compromise confidentiality that no disclosure certificate would fix.

Industry’s narrower ask

The technology industry body Nasscom took a different tack in its submission, broadly endorsing the framework while pressing for definitional clarity. It asked the Committee to define undefined terms the draft relies on — including how “high-risk” AI use should be classified — and to map that classification explicitly onto the permitted-use list in Regulation 19, so vendors and courts aren’t left to guess case-by-case. Nasscom also flagged that Regulation 38(2), which contemplates technical audits of AI systems used in courts, could be read to require access to source code or underlying algorithms — an expectation that would not survive contact with most third-party AI vendors, who routinely restrict that kind of access even to their institutional customers.

Where things stand

As of this writing, the picture is unresolved by design: the comment window closed on July 15, and the Committee has not yet published a final version or a notification date. That leaves India’s courts, and the lawyers who practice before them, in an unusual holding pattern — a widely discussed rulebook for AI in litigation that has no legal force yet, but is already shaping how firms think about disclosure, verification workflows, and AI tool procurement ahead of formal adoption.

What it means for lawyers and litigants

Whatever the Committee ultimately decides on Regulations 43(3) and 43(4), the substance of the disagreement is unlikely to disappear. Even SCAORA’s alternative — a certification of personal verification — assumes lawyers using AI tools are checking citations and holdings before filing, which is the practical safeguard regulators actually care about; India’s courts have already seen an AI-fabricated precedent reach a trial court order, with the Supreme Court warning that relying on non-existent, AI-invented judgments can amount to professional misconduct. For law firms, the sensible position is to treat verification as the baseline obligation regardless of which version of the rule survives, and to start documenting internal AI-use and citation-check workflows now, since some form of accountability requirement — disclosure, certification, or both — is coming to every Indian court once the apex body is constituted and implementation dates are notified. For litigants, the practical stakes are lower but real: a firm’s AI-verification practices are quickly becoming a legitimate diligence question when choosing counsel for high-stakes matters, particularly where citation accuracy and confidentiality of sensitive case facts are both in play.

Frequently Asked Questions

Has the Supreme Court’s draft AI regulation been finalised?

No. The AI Committee released the draft on June 3, 2026 for public comment, extended the comment deadline once to July 15, 2026, and as of now has not published a final version or a notification date bringing it into force.

What exactly must lawyers disclose under the draft rules?

Under draft Regulations 43(3) and 43(4), an advocate must certify “the AI-assisted character” of a pleading, document, or piece of evidence whenever an AI tool was used in preparing it, disclosing the nature and extent of that assistance at the time of filing.

Why is SCAORA opposed to the disclosure requirement?

SCAORA argues advocates are already personally accountable for every filing’s accuracy under the Advocates Act, 1961 and the Supreme Court Rules, 2013, making a separate AI-specific disclosure duplicative, “unjustified,” and potentially prejudicial to a client’s case or a judge’s perception of it.

Does the draft let AI decide cases or set bail?

No. Draft Regulation 20 expressly bars AI from independently adjudicating cases, determining bail eligibility, assessing witness credibility, or monitoring judges, advocates, or litigants without separate legal authorisation, and bars opaque AI systems wherever legal rights or personal liberty are at stake.

What AI tools are already used in Indian courts today?

Tools such as SUPACE (for judicial research assistance), SUVAS and SuSahayak (translation aids), and AI-enabled e-filing systems are already in use in pilot form across various courts, which SCAORA cited in asking for phased audits of existing deployments rather than a blanket new disclosure mandate for future use.

What happens next?

The AI Committee is expected to review the comments received (from SCAORA, Nasscom, individual judges, and other stakeholders), revise the draft, and eventually notify final regulations along with implementation timelines for the Supreme Court and each High Court — though no date for that has been announced.

Sources