Recusal's Double Standard: Too easy for Judges, too soft for Regulators

India runs 2 recusal regimes, and both have failed, in opposite directions. Judges increasingly step back from cases they are entitled to hear, and expected to hear, sometimes so readily that recusal looks less like restraint than a convenient exit from an awkward dispute; the same judges, in other matters, refuse to step back when every signal says they should. Regulators have had the mirror-image problem.

Until June 2026, the rules governing SEBI's own Chairperson and Whole-time Members amounted to a voluntary 5-page code with no penalty for breach, while the Board's own employees, several ranks below, answered to a proper, enforceable Service Regulation. The mismatch runs the wrong way: the top of each institution has been less bound than the bottom. That is worth setting out case by case.

India has no general statute on judicial recusal. The nearest thing is the 1997 Restatement of the Values of Judicial Life, aspirational principles the Supreme Court adopted for itself, binding on no one. What India has instead is case law going back to Manak Lal v Dr Prem Chand Singhvi (1957), establishing that a litigant need only show a reasonable apprehension of bias, not actual bias, to ask for the recusal of a decision-maker. There, a Bar Council tribunal's own Chairman had previously appeared against the accused advocate's clients; the Supreme Court held this to be disqualifying on its face, while separately holding that the advocate had forfeited the right to object by sitting on it until an adverse report came through, rather than raising it when he first learned the facts. Even the foundational judgment built an escape hatch into the doctrine meant to police bias.

That flexibility cuts both ways, sometimes within one judge's own record. As far back as 2009, Justice S.H. Kapadia recused from a petition on Vedanta's open offer for additional Sesa Goa shares, having disclosed he held shares in Sterlite Industries, Vedanta's sister company. Yet in an earlier matter where Sterlite was itself the party, over bauxite mining in the Niyamgiri hills, Kapadia had disclosed the identical shareholding, drawn no objection, and continued hearing that case to judgment. Same shares, same judge, opposite outcomes, turning only on whether anyone thought to object.

Reluctance can be just as costly as excess caution, and P.K. Ghosh v J.G. Rajput (1995) shows why. Justice B.J. Shethna of the Gujarat High Court had, before elevation, appeared as advocate for a municipal employee challenging his suspension by the Ahmedabad Municipal Corporation. Years later, now a judge, he heard a contempt petition arising from that very dispute. Counsel for the Corporation objected, on the returnable date and twice more after, that he should not sit on a case concerning a former client. He declined each time, and the Bench he sat on granted an interim stay in that client's favour. The Supreme Court quashed the order, holding that "justice should not only be done but it must also be seen to be done," and that recusal is owed not because a judge is likely to be actually swayed, but because his presence is "likely to give rise to a reasonable apprehension in the mind of the litigant" that the judge's mind, "maybe subconsciously," has been influenced.

Ranjit Thakur v Union of India (1987) takes the same doctrine outside the ordinary courts, to a summary court-martial. An army signalman, already serving a sentence his commanding officer had imposed, refused a later order to eat food, and was tried the very next day by a court-martial on which that same officer sat in judgment. The Supreme Court struck the proceedings down twice over: for non-compliance with Section 130 of the Army Act, which requires an accused be asked whether he objects to any officer on the Bench, a lapse going "to the root of the jurisdiction" regardless of actual prejudice; and because an officer who had punished the accused, and been complained against by him, could not then sit in judgment on him, rendering the proceeding "coram non-judice." The test, the Court said, "is not to look at his own mind and ask himself, however, honestly, 'Am I biased?'; but to look at the mind of the party before him," the same test P.K. Ghosh would apply 8 years later.

Given a doctrine this developed, one might expect the Court to welcome a chance to put it on firmer footing. It has done the opposite. In Chandraprabha v Union of India (May 2025), Justices Abhay S. Oka and Ujjal Bhuyan dismissed a petition asking the Court to frame binding guidelines on recusal under its own Article 142 power, holding flatly that "recusal is a matter of discretion of the judge." Asked directly to close the gap, the Court chose to keep it open.

What that gap costs in practice is visible in a single litigation. The challenge to the law removing the Chief Justice of India from the panel selecting Election Commissioners has already claimed 2 Chief Justices: Sanjiv Khanna, in December 2024, and Surya Kant, in March 2026, who directed that the case goes to judges outside the line of succession because it concerns the Chief Justice's own role in the selection panel. Yet when a different Bench delivered a split verdict on a threshold question in that case on 23 September 2026, it emerged 2 days later, through media reporting rather than anything disclosed in court, that one of its 2 judges had a son serving as the Election Commission's own standing counsel before the Madhya Pradesh High Court since mid-2024. A plea has since sought recall of the verdict. The point worth pressing is not a verdict on any individual judge, but the policy question the case leaves open: should a judge's family member holding a standing brief for a party before him trigger disclosure as a matter of course, whatever the outcome? A published, prospective standard would settle that before a case is heard. As things stand, it surfaces only after, through a recall plea, if it surfaces at all.

A quieter version of the same problem runs through the long battle over NLSIU Bangalore's domicile reservation. 4 judges, Justices U.U. Lalit, Aniruddha Bose, Abdul Nazeer and Hrishikesh Roy, recused in succession between 2021 and 2023. Justice Lalit said he had earlier, as a lawyer, represented a member of the university's Governing Council; reports of the other 3 recusals record no reason. The case was finally disposed of in July 2024, on Karnataka's withdrawal rather than a verdict on merits, by a Bench that included Justice Pankaj Mithal, whose own official Supreme Court biography lists him as a nominee to the university's Executive Council. None of this suggests Justice Mithal was disqualified. It does show the limit of recusals without reasons: the public cannot tell what links the 4 judges had, or whether the Bench that finally disposed of the case shared them.

What these episodes add up to is not a judiciary that recuses too often or too rarely in any stable direction. It is one where the decision, either way, is made privately and reviewed by nobody but the judge who made it.

The regulatory story is the mirror image. When Hindenburg Research alleged in August 2024 that then SEBI Chairperson Madhabi Puri Buch and her husband held stakes in offshore entities tied to Adani-linked fund structures, SEBI's answer was that its internal mechanisms, including a "disclosure framework and provision for recusal," were adequate, and that the Chairperson had "recused herself in matters involving potential conflicts of interest." It did not say whether Adani was among those matters. Buch's own statement named Blackstone as being on her recusal list. When the question reached the Lokpal in 2025, her answer on Adani was not that she had recused but that there was nothing to recuse from: the fund she had invested in held no Adani shares, and no Adani file had been put up to her. The Lokpal accepted that and dismissed the complaints against her. The episode exposed not a breach of SEBI's rules so much as how little those rules amounted to. The governing instrument, the SEBI Code on Conflict of Interests for Members of Board, 2008, asked a Member who spotted a conflict to disclose it to the Chairman, who, in practice, alone decided if it was genuine; disclosures were kept confidential and generally filed with the Secretary to the Board in a sealed cover, so that even the Secretary never saw their contents. Section 7A of the SEBI Act and the SEBI (Procedure for Board Meetings) Regulations, 2001 added only a duty to disclose an interest in a matter before the Board and abstain from it.

Set against this, SEBI's own employees, since 2001, had answered to the SEBI (Employees' Service) Regulations: enforceable subordinate legislation barring speculative investment, conferring "insider" status, and requiring annual asset declarations, backed by disciplinary penalties the 2008 Code never extended to the Members those employees served.

SEBI's Board confronted this gap only once the Buch controversy forced its hand. At its meeting of 24 March 2025, the first under new Chairman Tuhin Kanta Pandey, it set up a High-Level Committee under former Chief Vigilance Commissioner Pratyush Sinha, with Injeti Srinivas as Vice-Chairman and Uday Kotak, G. Mahalingam, Sarit Jafa and Professor R. Narayanaswamy as Members. Its report, submitted on 10 November 2025, did not pull punches: the 2008 Code cited no statutory source, its legal validity had never once been tested in court, and it amounted, in the Committee's own words, to "nothing more than a voluntary statement of expected conduct," carrying no penalty at all, unlike the fully enforceable ESR. It noted, almost in passing, that since January 2018 the Chairman and Whole-time Members had voluntarily filed immovable property returns in the format prescribed for All India Service officers, and observed the trading restrictions that apply to employees, an informal practice resting on nothing more than a future Board's willingness to keep it up.

The Board's own first reaction to this report was to shelve it. At its meeting of 17 December 2025, it declined to adopt the recommendations: some members raised constitutional doubts about whether SEBI could even bind officials whose service conditions are fixed by the Central Government's 1992 Rules, others raised privacy concerns over public asset disclosure, and the matter was deferred.

It returned to the question at its 213^th meeting on 23 March 2026, and this time accepted most of what the Committee proposed: the Chairman and Whole-time Members were brought within the "insider" definition, investment in the pooled-fund products of any one regulated entity was capped at 25% of a portfolio, and public disclosure of immovable property was extended to Executive Directors and Chief General Managers. Tellingly, it stopped short of 2 further proposals, a standalone regulation for Board Members and an independent oversight mechanism, referring both to the Central Government, since only Government, as appointing authority, could put them into effect.

On 19 June 2026, the Board adopted the 17-page Code of Conduct for Members of SEBI, 2026, covering Whole-time and Part-time Members alike, a genuine advance on paper: disclosures now go to a newly created Office of Ethics and Compliance rather than to a sealed cover, a digital system logs recusals with a named count against each Member, Executive Director and Chief General Manager published yearly, and public complaints are screened by the Office before reaching the Chairman or Board. Yet the mechanism for resolving a genuine doubt is in substance what it was in 2008: a question about the Chairperson's own conflict goes to the "Board excluding Chairperson," a question about any other Whole-time Member goes to the Chairperson alone. The same small, self-referential circle still decides its own members' conflicts, with one more screening step in front of it.

Nowhere is the limit of what the Board could do alone clearer than in the Code's treatment of what happens after a Member leaves. Clause 12(3) bars a Whole-time Member, for 2 years, from appearing before or against SEBI on anyone else's behalf, a restriction within the Board's own regulatory reach. Clause 12(1), on post-retirement employment itself, does not attempt to change the rule at all: it simply defers to "the SEBI (Terms and Conditions of Service of Chairman and Members) Rules, 1992," the same 1-year cooling-off that is shorter than the 2 years at the IFSCA and the PFRDA. The preamble gives the game away, recording that the Code was "adopted voluntarily" by the Board, the identical status the discredited 2008 Code carried, because the terms of service of its Chairperson and Members are fixed by Government rules, not by SEBI. Employees, by contrast, answer to SEBI's own regulations, which carry the force of law: on 11 July 2026, the SEBI (Employees' Service) (Amendment) Regulations, 2026 came into force, moving the staff's 2-year bar on representing clients before SEBI after leaving out of an internal circular and into the regulations themselves, adding mandatory disclosure of employment negotiations, and placing both on exactly the statutory footing the Board's own Members still lack.

The instrument capable of closing that gap is already before Parliament. The Securities Markets Code, 2025, examined by the Standing Committee on Finance (2025-26) of the 18th Lok Sabha, restates the Chairperson's and Members' terms of service in Clause 5, and Clause 5(6) re-enacts the 1-year cooling-off of Rule 3(4), even though the IFSCA and the PFRDA both prescribe 2 years. Explaining the choice, the Secretary, Ministry of Finance, told the Committee that 2 years had originally been on the table and was cut to 1 "in order to attract talent" and "preserve the attractiveness of serving on the SEBI Board." The Committee did not buy it, noting SEBI's "wide-ranging regulatory, supervisory, investigative and enforcement powers," and recommended Clause 5(6) be restored to 2 years.

The convergence is worth sitting with. 2 separate exercises, SEBI's own executive-appointed Committee and Parliament's own Standing Committee, with no apparent coordination between them, landed on the identical number. Whether that survives into the statute, or merely becomes a second report reaching a finding the Ministry had already declined once, will be clear only once the Bill completes its remaining stages.

Neither story is really about excess or deficiency of recusal. Both are about the absence of anything that forces a reason into the open. A judiciary with no codified standard, and no apparent wish for one even when directly asked, has too much discretion and no one to answer to for how it is used. A regulator whose most senior office was, until June 2026, governed by a confidential, unenforceable code had too little law where the risk of conflict was greatest. Codification would help close the second gap; SEBI's own Members could be placed on the same footing as the staff who answer to them, and there is no obvious reason it should wait.

But the more urgent correction, for courts and regulators alike, costs less: every recusal and every refusal to recuse where one was sought should come with a brief reason, put on the public record, so that an apprehension of bias can be tested by someone other than the one person least able to judge it.

Sumit Agrawal is the Managing Partner, Regstreet Law Advisors & former SEBI Officer.

M. Damodaran is the Chairperson, Excellence Enablers and former Chairman, SEBI, UTI & IDBI.

Views are personal.