[This is a guest post by Kartik Kalra.]
“I will not recuse”, writes Swarna Kanta Sharma, J., in bold letters, in her judgment dated 20 April 2026, holding that the submissions of six applicants seeking her recusal – on grounds including her prior adjudication of identical legal questions in separate proceedings – amounted to an attack on her “dignity” and “integrity”, a “personal attack” against her, an “attack on the institution” of the judiciary, and an attempt to “lower the judicial process itself by raising unfounded allegations” [¶¶5, 11, 109, 192].
The case before her was a revision petition against a judgment of a Special Court issued in February, which rejected the Central Bureau of Investigation’s (“CBI”) allegations of conspiracy, bribery, and destruction of evidence against twenty-three accused, including former Delhi Chief Minister Arvind Kejriwal. The CBI’s case against Mr. Kejriwal and other members or affiliates of the Aam Aadmi Party (“AAP”) rested on three prongs:
(a) First, that members of the AAP, the then party in power in Delhi, received bribes to alter the Delhi Excise Policy 2021-22, and in return, enacted a policy that increased profit margins of liquor wholesalers from 5% to 12%;
(b) Second, that AAP, under Mr. Kejriwal’s leadership, extracted kickbacks from those liquor wholesalers, who paid these sums out of their enhanced margins; and
(c) And third, that AAP used these kickback amounts to finance its election campaign in Goa. This last prong overlaps with the case of the Enforcement Directorate, which considers this alleged election spending as the use of “proceeds of crime” u/s 3 of the Prevention of Money Laundering Act 2002 (“PMLA”).
This theory, the Special Court found, was “economically illiterate” and “legally unsustainable” [¶357]. It held that the 2021-22 policy was, in fact, more economically burdensome for wholesalers than the earlier regime, because it prevented them from recovering several costs associated with the sale of liquor to retailers, including freight, handling, insurance, and warehousing, which previous policies permitted [¶352]. Given this finding that wholesalers did not hold the economic surplus to finance kickbacks in the first place, together with the CBI’s heavy reliance on uncorroborated approver statements, and the fact that the policy itself lay in the domain of lawmaking and the sums in question were legally permitted earnings of wholesalers, the Special Court held that no case could be suspected against any accused [¶1077]. Accordingly, it discharged all of them [¶1085].
It is, however, this very theory that Sharma, J. has fully endorsed thrice. In Arvind Kejriwal v Directorate of Enforcement, she held that this same broad set of facts “point[ed] towards [Kejriwal’s] guilt” [¶114]; in Manish Sisodia v Directorate of Enforcement, she found that Mr. Manish Sisodia, a co-accused, had “commit[ed] the offence of money laundering” [¶151]; and in Sanjay Singh v Union of India, she found Mr. Sanjay Singh, an AAP member, “prima facie guilty of offence under PMLA” [¶48]. This sharp divergence in the treatment of foundational factual questions formed the core of the case advanced by the six discharged individuals seeking Sharma, J.’s recusal. The Special Court held these facts incapable of raising even a suspicion of guilt to frame charges, yet Sharma, J. found them sufficient to conclude a verdict of likely guilt to authorise detention and hold arrests valid. Since propositions wholly contrary to those noted by the Special Court had already been endorsed by Sharma, J., a reasonable apprehension of her retaining her original views could be inferred by a fair-minded observer, warranting her recusal.
In her judgment, however, she rejected each ground advanced to establish a reasonable apprehension of bias, including – in addition to her previous findings of the discharged individuals’ likely guilt – her earlier finding that the Special Court’s order of discharge “prima facie appear[ed] erroneous” [¶8], her presence in events of political-legal bodies, as well as issues of conflicts of interest arising from her family members’ empanelment with the government. She described these concerns as “absurd”, being “illusions and speculations”, and held that instead of being grounded in any issue concerning her, they were an issue “in the mind of [Kejriwal]” himself [¶¶109, 234, 169].
Institutional Catastrophe as a Basis for ̉Refusing Recusal
A reader would note a common thread of reasoning across each of the eight grounds of recusal that Sharma, J. rejects: if accepted as valid grounds for recusal, their broader institutional consequences would, in her view, be disastrous for the judicial system. On that basis, she treats them as incapable of generating a reasonable apprehension of bias, and therefore refuses to recuse. The same reasoning is offered to resist each of the eight grounds of recusal, four of which are as follows:
- Ground 1 – Apprehension of bias arising from a previous order finding the Special Court’s judgment prima facie erroneous: “If the argument advanced by the applicants is accepted, no court would ever be able to grant any interim relief or pass any order of stay.” [¶78]
- Ground 3 – Apprehension of bias arising from earlier judgments on PMLA arrest and bail: “If such an argument were to be accepted, it would mean that no Judge who has ever decided a bail application, anticipatory bail application, or challenge to arrest in a criminal case would thereafter be competent to hear proceedings arising from the same FIR.” [¶132]
- Ground 6 – Apprehension of bias arising from the judge’s participation in events of a legal-political body: “Judges, by this standard, would be forced to withdraw from public legal discourse and from sharing their experiences and knowledge with younger members of the Bar and others.” [¶164]
- Ground 7 – Apprehension of bias arising from professional engagements of relatives: If this argument is accepted, “then a large part of the judiciary, from the District Courts to the highest Court, would have to recuse from hearing such matters.” [¶181]. Similar reasoning is used with Grounds 2, 4, 5 and 8 [¶¶112, 145, 149, 251].
It is, therefore, Sharma, J.’s perceived undesirability of the systemic consequences of recusing on these grounds that leads her to characterise the apprehension of bias as unreasonable. Kejriwal’s submissions are abstracted to their most general form and then rejected as incompatible with the ordinary functioning of courts: judges must be able to hear related matters, make candid prima facie observations, participate in public legal life, and have family members pursue careers in the legal profession. Kejriwal, in challenging this state of affairs, must be challenging the judicial institution itself, and a refusal to recuse would be essential to “defend [her]self” and the “collective integrity of the institution of the judiciary” [¶253-4].
That, however, was not what Kejriwal – or any other discharged individual – argued. By abstracting their submissions to the most general level, Sharma J. transforms the enquiry from assessing a reasonable apprehension in this case, to one where extensive policy considerations of the general consequences of a recusal govern whether it should occur. In this shift, the question that is ultimately answered is whether a recusal is desirable considering its consequent impact on the general public perception of the judiciary, with highly remote, unforeseeable, and improbable implications of a recusal – such as the installation of a bar on judges’ issuance of prima facie remarks or on their hearing of cases involving overlapping facts – being given centrality.
It is settled that the individual from whose perspective an apprehension of bias must be assessed is a “reasonable and fair-minded” individual – the “objective observer” – and not the judge themselves. In the instant case, however, the set of perspectives that constitute reasons for Sharma, J.’s conclusion of the unreasonableness of apprehending bias are squarely her own. A reasonable and fair-minded individual, it is submitted, would not accord centrality to a singular set of considerations emanating from remote and unforeseeable impacts on the judiciary’s reputation if – in the future – the grounds urged by Kejrwal become common grounds for recusal. That individual would, instead, answer the simple, fact-specific, question that is posed: do Sharma, J.’s previous factual findings of Kejriwal’s (and other co-accused persons’) likely guilt in validating their arrest and declining bail give rise to a reasonable apprehension that she may retain those views in now assessing their possible guilt for discharge?
The answer, in my view, was squarely straightforward, and the story of heavens falling upon a recusal merely deflects from that question. Deciding this case, according to Sharma, J., was necessary because her “duty demand[ed] that [the] case be decided”, and refraining from doing so would constitute “abdication of duty” [¶¶246, 258]. The rhetoric of protecting the judiciary from various forms of remote harm, however, eclipses the obvious reality that she owes no duty to decide this case, that it is not necessary – in terms of the absence of effective alternatives to her being the judge – for her to hear this revision, and that no prejudice would occur to any party does she recuse. Upon the existence of a reasonable apprehension of bias, the “doctrine of necessity”, noted in a recent decision, requires for the judge to continue hearing a case only if there is no other person competent to decide it or if they are essential to constitute a quorum [¶86], neither of which apply here. There are forty-three other sitting judges of the Delhi High Court, with six judges specifically allotted the roster of hearing criminal revision petitions filed in 2026, any among whom may hear this case.
Importantly, in Madhyamam Broadcasting v Union of India, the Supreme Court found that infringements of natural justice, which includes the rule against bias, are justified only upon satisfying principles of proportionality [¶57]. One ingredient of proportionality, as noted in Electoral Bonds, is that the applicant need only discharge a prima facie burden of showing infringement, after which the burden shifts to the party defending it [¶44]. Applied here, once the discharged individuals show a prima facie case of apprehending bias, their burden ought to have been treated as satisfied, and recusal would ordinarily follow unless that apprehension was successfully rebutted. In the limited part of Sharma, J.’s judgment that does address the issue of her existing responses to identical legal questions, there is little available to conclude how a fair-minded observer would not hold an apprehension of bias.
On Factual Prejudgment and Apprehended Bias
Sharma, J. offers three reasons for rejecting the discharged individuals’ submission that her earlier judgments give rise to a reasonable apprehension of bias. First, that it was not on her own volition that she recorded the findings of their likely guilt, for the PMLA, u/ss 19 and 45, itself requires judges to come to these findings, which, in any case, were only prima facie in nature [¶120-1]. Second, that a discharge order was legally distinct from orders on arrest or bail, and that she had not heard any case for discharge till yet [¶124]. And third, she cites precedent from the Supreme Court discouraging recusals on grounds of subject-matter bias [¶122]. None among these would dispel the apprehension of bias in the mind of an objective observer.
First, the question to be answered for discharge under criminal law, pursuant to well-settled case-law, is whether a “prima facie case against the accused has been made out” [¶17]. Simultaneously, the question to be answered in a case challenging arrest u/s 19 or for bail u/s 45 of the PMLA – the provisions under which Sharma, J. has issued three judgments on the same facts – is whether there is “reason to believe that [a] person [is] guilty” and whether there exist “reasonable grounds for believing that [one] is guilty” respectively. Syntactic differences apart, all three enquiries hold a similar functional meaning, requiring the judge to answer the same question of their belief in the accused’s possible or likely guilt. It may, in fact, be argued that the threshold to find a PMLA arrest or detention valid is higher than to frame charges, for the former two require a judge’s belief in one’s guilt, while the latter requires only a prima facie case.
Accordingly, Sharma, J. has either answered the same functional question – that of the discharged individuals’ possible guilt – against them, or she has answered a question of higher threshold, that of discharged individuals’ likely guilt under the PMLA, also against them. The fact that these questions arise from different statutory provisions is of little logical significance, for a fair-minded observer would apprehend bias from the substance of Sharma, J.’s findings, not their precise legal source. It is, in fact, highly unlikely that Sharma, J. could previously have adjudicated an order for discharge arising from the same legal provision. The case concerns a criminal conspiracy involving twenty-three co-accused, whose discharge was necessarily decided as one case, simultaneously, by the Special Court because the offence of conspiracy is, definitionally, an offence committed by a group. Once that common discharge order was issued, its revision would also be heard together, as one case, by the High Court. The legal source of Sharma, J.’s earlier findings, therefore, would necessarily be different, but this difference would be inconsequential to an objective observer’s apprehension of bias.
And second, Sharma, J. cites the Supreme Court’s holding in Indore Development Authority v. Manohar Lal to hold that a “previous judgment cannot constitute bias…[or]…raise a reasonable apprehension of bias” [¶122]. Indore concerned a request for the recusal of Mishra, J., who had earlier been part of a three-judge bench deciding the interpretation of a provision that was then being reconsidered by a five-judge bench, of which he was also a member. In refusing recusal, Mishra, J. held that judges remain open to reconsidering the law, and that “the entire judicial system exists for refining” the law [¶31]. It does, indeed, occur that judges – when sitting in larger benches – reconsider their prior holdings and conclude their incorrectness. Khanna, J., in his separate opinion in Interplay Between Arbitration Agreements, found his previous views in Vidya Drolia warranting modification [¶41], and so did Untwalia, J. in Madhu Limaye v State of Maharashtra.
What is happening in the instant case, however, is markedly different from the phenomenon of judges’ views on the law evolving with time or being revised with further reflection. This case concerns straightforward findings on fact which have already been reached, and an apprehension of bias arising from the necessary re-answering of essentially the same factual question that Sharma, J. has thrice answered. Indore – the very case Sharma, J. relies on to refuse recusing – notes the distinction between concerns of bias arising from “general questions of law [and] policy” and the “prejudging of facts specifically relating to a party”, with the latter necessitating recusal [¶27]. This distinction, and the necessity to recuse in cases concerning a judge’s previous factual findings on identical questions, is well-established in precedent [¶30]. For comparative context, the Australian High Court, in its decision in Livesey v New South Wales Bar Association, found that a judge’s expression of “clear views about a question of fact which constitute[d] a live and significant issue in [a] subsequent case” necessitated recusal [pg. 299-300]. Similarly, the European Court of Human Rights, in an analogous setting where a judge expressed a (statutorily mandated) “particularly confirmed suspicion” of guilt against an accused to authorize their pre-trial detention, found that a trial by the same judge – where a similar question, i.e., the accused’s guilt is adjudicated – would violate the guarantee to an impartial tribunal [¶52]. Clearly, therefore, Sharma, J.’s extensive factual findings of the likely guilt of multiple among the discharged individuals, arising from the same common set of facts that the Special Court found giving rise to an “economically illiterate” case, gives rise to a reasonable apprehension. Any “real ground for doubt”, it has been noted, must be resolved in favour of recusal [¶18].
In holding that bias cannot be inferred because judges are capable of approaching similar questions with an open mind [¶134], Sharma, J. transforms the standard from a reasonable apprehension to actual bias, and the latter does not exist, for she is, like any judge, capable of approaching this case without any predetermined opinion giving way [¶206]. Sharma, J.’s own individual ability to detach herself from her previous findings, it is submitted, is wholly irrelevant to the decision to recuse. Recusal does not turn on a judge’s assurances and abilities to remain impartial, for the question of the physical, psychological state of bias in an individual’s mind is incapable of determination. This is why Indian – and global – jurisprudence on the rule against bias does not ask the question of a judge’s ability to decide without bias, but rather whether a fair-minded individual would apprehend bias. I submit that any such individual would, given functionally similar enquiries for discharge and PMLA arrest and bail, and her extensive findings of the discharged individuals’ guilt for the latter through three judgments.
The Judgment Refusing Recusal as a Further Source of Apprehended Bias
Importantly, the judgment concludes with Sharma, J.’s observation that she “carries no bitterness” against discharged individuals for moving the recusal application, and that she would continue deciding the case “totally uninfluenced and unaffected by these recusal applications” [¶264-5].
The difficulty, however, lies in the judgment’s repeated use of strong language in characterising Kejriwal’s submissions. It states, for instance, that he used recusal as a “weapon to intimidate a judge” [¶166], that he seeks to “damage an institution by making insinuations against a Judge” [¶229], that he seeks to make it “normal to intimidate judges” [¶166], that his arguments “arrived on [her] table with aspersions, insinuations, and doubts cast on [her] fairness and impartiality” [¶252], and that he has “lower[ed] the judicial process” [¶192]. These observations may themselves constitute additional reasons for apprehending bias.
The use of strong language in a judgment, particularly when its use is wholly unnecessary to decide the legal question before a court, has been found to give rise to apprehensions of bias (see the Supreme Court’s decision in Kanaklata v NCT of Delhi, ¶5 and Tripura High Court’s decision in Prosecution v Sumit Banik, ¶32). Since the instant recusal applications could have been decided without issuing any such remarks, purely from the perspective of what an objective observer would think of Sharma, J.’s factual prejudgment over an identical legal issue, each of these remarks were wholly unnecessary, and may now serve as independent grounds to seek her recusal.
Conclusion
Accordingly, I submit that an objective observer would reasonably apprehend bias if Sharma, J. were to hear the revision against discharge. The suggestion that Kejriwal’s submissions would cause institutional catastrophe deflects from the real question of an objective observer’s apprehension arising from factual prejudgment, and allows one set of concerns – particularly from Sharma, J.’s own perspective – to substitute what an objective observer would think. Precedent discouraging recusals owing to judges’ revisiting of their prior holdings misrepresents the case as one of a general question of law or interpretation, and the authorities that Sharma, J. relies on for this proposition themselves recognise the distinction between prior findings on questions of law and prior findings on questions of fact, with the latter necessitating recusal. The judgment’s use of strong language against Kejrwial, despite the fact that the recusal application could have been decided by answering a narrow and straightforward legal question, may itself give rise to new grounds for apprehending bias and seeking recusal.