Guest Post: Illegally arrested, Legally Detained - The Paradox of Habeas Corpus

(This is a guest post by Bipul Kumar and Vaneet Kumar)

Recently, the petitioner, an accused under murder case, in Neeraj v. State of UP (“Neeraj”), filed a writ of habeas corpus in Allahabad High Court arguing that his remand is illegal since he was not informed of the grounds of his arrest. However, Neeraj vide its order dated 27 May 2026, rejected the petition by holding that the post-cognizance remand cannot be vitiated because of illegal arrest (para 88). Moreover, it also declared Mihir Rajesh Shah v. State of Maharashtra line of judgments that declared arrest and subsequent remand illegal on non-communication of grounds as per incuriam, for conflicting with Kanu Sanyal v. DM, Darjeeling line of judgments (paras 87). The Court, while reaching these conclusions, also introduced two limitation ‘points’ on the availability of habeas corpus petition against arrest and remand: (a) taking of cognizance, and (b) the rejection of bail application. This article upon critically engaging with Neeraj argues that while the High Court correctly limits the remedy to pre-cognizance detention, its holding of per incuriam and its second limitation on habeas corpus are legally unsustainable.

Is Mihir Rajesh Shah Really Per Incuriam?

Neeraj concluded that Mihir Rajesh Shah and like “are not binding precedents and hit by the principles of stare decisis” (that is, per incuriam), without providing any reason provided alongside this conclusion (paras 86-87). This conclusion invites two lines of enquiry: (a) Whether the High Court can declare a precedent as per incuriam? And (b) Whether the conclusion reached by Neeraj is correct, irrespective of its competence? On the first line of enquiry, the answer can be found in Sundeep Kumar Bafna v. State of Maharashtra, where the Supreme Court held that the all courts, including High Courts, can declare a decision to be per incuriam; however, such declaration should be made with extreme caution and circumspection (para 16). Therefore, Allahabad High Court, via Neeraj, is competent to declare Mihir Rajesh Shah per incuriam.

As per Sundeep Kumar Bafna, “a decision or judgment can also be per incuriam if it is not possible to reconcile its ratio with that of a previously pronounced judgment” (para 15). There should be a “glaring case of obtrusive omission”, as per Bajaj Alliance General Insurance Company v. Rambha Devi. In simple terms, if the overlooked binding precedent alters the conclusion of the judgment, such judgment is per incuriam. It is the assertion of Neeraj that Mihir Rajesh Shah line of decisions “do not appear to have considered the aforesaid earlier pronouncements of the Apex Court” (that is, Kanu Sanyal line of judgments) (para 77). However, Neeraj proceed to declare the former line of judgments as per incuriam without so much as asserting the irreconcilability of the two lines of judgments. This renders its conclusion doubly flawed, falling short of both the legal test enunciated in Bajaj Alliance and the salutary caution prescribed in Sundeep Kumar Bafna. Unlike Neeraj, it is our assertion that these two lines of judgments are not competing at all and, even if deemed to be so, are readily reconcilable. To enunciate, while the Mihir Rajesh Shah line of judgments deal with the post-arrest remand orders issued under section 187 of the Bharatiya Nagarik Suraksha Sanhita (“BNSS”) (or pari materia section 167 of the Code of Criminal Procedure, 1973), the Kanu Sanyal line of judgments deal with the preventive detention or remand orders issued under sections other than section 187 of the BNSS. Since the basis of remand is different under the two lines of judgments, they were never at loggerheads.

First Limitation ‘Point’

Although there is no limitation ‘period’ for filing a writ petition for enforcing the fundamental rights, especially under Article 22(1) of the Constitution (it mandates communication of grounds of arrest), Neeraj introduces limitation ‘points’ for this remedy. We shall discuss these two ‘points’ at length in this and following section. First, as mentioned above, Neeraj postulates that cognizance is a cut-off point for filing a habeas corpus writ under Mihir Rajesh Shah regime. This conclusion, in our opinion, is but obvious, and flows from multiple reasons which we proceed to address in turn. But before turning to these reasons, it is important to distinguish between the pre-cognizance and post-cognizance remand orders. While the remand orders under section 187 of the BNSS are pre-cognizance orders, the remand orders under sections 232 and 346 of the BNSS (or pari materia sections 209 and 309 of the Code of Criminal Procedure, 1973, respectively) are made after cognizance and hence, called post-cognizance orders. Neeraj postulate the reason as follows: “The order taking cognizance is a judicial order and stands on a higher footing than an order of remand passed under [section 187 of the BNSS], which is merely a pre-investigation judicial order” (para 75).

This quote provides us with first reason that the post-cognizance remand is based on separate considerations (that is, investigative material or chargesheet) that pre-cognizance remand (that is, pre-investigation material), making it legally untenable to adjudge the legality of the former detention on the basis of legality of the latter detention. The second reason, albeit absent from Neeraj, is straightforward: while pre-cognizance remand cannot be made but for arrest, the post-cognizance remand can be made independent of arrest. Therefore, the legality of arrest cannot impose as a litmus test for post-cognizance test. Thirdly, doing it otherwise would have violated the well-established and accepted ‘rule of return’. Fourthly, the factual matrix underlying Mihir Rajesh Shah regime concerned itself exclusively with pre-cognizance remand orders. The extension of that regime to post-cognizance remand orders would therefore be contrary to the doctrine of stare decisis, which demands that precedents be applied only to cases arising from analogous facts and circumstances. Therefore, cognisance as the ‘point’ of limitation for writ of habeas corpus can be founded well-established in law.

Before moving to the second ‘point’, it is pertinent to discuss the murkiness that lies beneath the superficially clear limitation ‘point’ of cognizance. Suppose X is illegally arrested and remanded under section 187 of the BNSS. The police subsequently file a charge-sheet under Section 193 of the BNSS; however, as is invariably the case in practice, there ensues a period of delay before the Court takes cognizance. Now, the writ of habeas corpus filed by X is heard at this stage. Should X be released? Now, as per the discussion up till now, the answer might appear yes. However, the answer is not as straightforward as it seems. Surprisingly and unfortunately, there is no provision that governs the detention during the period between charge-sheet and cognizance because while section 187 of the BNSS, as per Sanjay Dutt v. State, ceases to operate upon filing of chargesheet, sections 232 and 346 of the BNSS, as per State of UP v. Laxmi Brahman, cannot be applied until cognizance is taken.

Despite this peculiar situation, Neeraj, without providing any reason, declares that remand under section 187 of the BNSS “remains operative only till cognizance of the offence” (para 74). This inference despite treating X’s remand and pre-cognizance alike is hit by the doctrine of per incuriam as it ignores Sanjay Dutt. Contrarily, the Delhi High Court in Nitin Nagpal v. State declared X’s remand patently illegal (para 18). Therefore, the X’s writ of habeas corpus under Mihir Rajesh Shah regime will succeed in either scenario; however, X does not even need to allege ignorance of grounds, if Nitin Nagpal is considered. Conversely, the Allahabad High Court in Sudhakar Rastogi v. State of UP, by ruling that X’s remand is governed by section 346 of the BNSS, ends up classifying it as post-cognizance remand. This classification renders the X’s writ infructuous due to first limitation ‘point’ of Neeraj. Therefore, writ filed by an accused, who is in position similar to that of X, will receive different answers by different benches or Courts, even though we believe that Delhi High Court’s opinion is correct because the Courts, as per Natabar Parida v. State Of Orissa, do not have inherent power to detain.

Second Limitation ‘Point’

On the second limitation ‘point’, Neeraj reasons that subsequent bench should not entertain habeas corpus writ if the High Court or Supreme Court has rejected the bail application because doing otherwise “will amount to appeal/review of the judgment passed by the bench deciding bail application” (para 88(ii)). Even if this reasoning seems palatable, we contend that it is incorrect representation of law because while bail application contends that the accused satisfies the triple test and poses no risk of flight, tampering, or reoffending, a writ of habeas corpus challenges the legal validity of the detention at its very root. Furthermore, the distinction also lies in the judicial approach: while bail is discretionary and admits of no straitjacket formula, the writ of habeas corpus is, in a sense, compulsory. This may appear an extreme proposition; however, what is meant is simply that under the Mihir Rajesh Shah regime, the writ operates on a near-mechanical basis — failure to communicate the grounds of arrest must necessarily result in release from pre-cognizance custody. If the grounds, purposes, and operation of two are inherently different, Neeraj ought not to consider one as an appeal or review of another, simply because both ends up in same outcome (that is, liberty of the accused).

Rather, the Courts have even entertained both the applications simultaneously and one after another. For example, Arvind Kejriwal (in CBI matter) filed both writ of habeas corpus and bail application, and the Delhi High Court vide two separate orders decided these two petitions (see Arvind Kejriwal v. CBI (I) and Arvind Kejriwal v. CBI (II)). For another example, Kejriwal (in ED matter) filed a bail application in the Special Court, after the Supreme Court in Arvind Kejriwal v. ED reserved the matter involving writ of habeas corpus. While it is conceded that neither of the examples exhibit a bail application filed after habeas corpus writ; however, these examples exemplify that the courts have treated both the petitions differently. Critics, like Neeraj, might resort to Eastern Coalfields Ltd. v. Dugal Kumar to argue that the accused should exercise writ at the earliest reasonable opportunity. Though we concede that this allows the accused to avail relief at the earliest opportunity, this in itself cannot act as a bar for a writ petition filed after the bail application is rejected, especially when first limitation ‘point’ exists.

Conclusion

Upon culminating all the arguments, it can be concluded that the Neeraj not only wrongfully declared the Mihir Rajesh Shah line of judgments as per incuriam but also provided bail rejection as a legally untenable limitation ‘point’ for filing writ petition of habeas corpus. Upon removing these observations from Neeraj, the correct legal position can be stated as follows: the accused can file a writ petition of habeas corpus for illegal arrest until the court takes cognizance and detains the accused under post-cognizance remand orders.

Furthermore, it is important to note that this analysis of Allahabad High Court’s judgment comes at a crucial time, especially when the Supreme Court in State of Meghalaya v. Sonam Raghuvanshi on 23 July 2026 created a distinction between non-service of grounds and insufficiency of reasons stated therein. In our opinion, it effectively limited the scope of Mihir Rajesh Shah to form (that is, service of grounds) and hinged the vitiation of arrest on substance of grounds (that is, prejudice caused due to insufficiency of reasons). This might give rise to mechanical act of furnishing grounds of arrest by the police, leaving the accused to prove the prejudice before the courts — ultimately, becoming a brewing ground for upcoming jurisprudence on the grounds of arrest and its intersection with Mihir Rajesh Shah. Interestingly, the apex Court while hearing this very matter, like Neeraj, orally mulled referring Mihir Rajesh Shah to a larger bench for being per incuriam. However, the Court, without delving into reasons, relied on Shah, leaving the question of per incuriam for a future occasion. This constant attack on Mihir Rajesh Shah also exposes the possibility of potential judicial conflict. Hence, this piece should not be confined to the vision of Neeraj, but rather be considered against the ever-expanding jurisprudence on ‘grounds of arrest’.

Lastly, as a parting note, it is important to recollect that, as per Vihaan Kumar v. State of Haryana (one of the judgments in Mihir Rajesh Shah regime), the illegal arrest violates fundamental rights enshrined under both Article 21 and 22. The gravity of violation is well reflected by Supreme Court’s statement: “Deprivation of liberty even for a single day is one day too many” in Arnab Manoranjan Goswami v. The State of Maharashtra. Considering the gravity of violation, we opine that the shift to post-cognizance remand should not leave the arrestee remediless; rather, the accused should be entitled to compensation for illegal detention. The Supreme Court in Bhim Singh v. State of J & K, a case dealing with illegal detention, provided compensation even though illegal detention has ended. Therefore, the Constitutional Courts like Bhim Singh should push for compensation, in addition to release, as a relief under the writ of habeas corpus to deter the state machinery from violating the accused’s fundamental rights.

Guest Post: Decoding the Doctrinal Shift in Section 223 BNSS (?)

(This is a guest post by Akshitha P. Hariharan)

The Bharatiya Nagarik Suraksha Sanhita, 2023 (“BNSS”) has replaced the Code of Criminal Procedure, 1973 (“CrPC”) with considerable fanfare about reform and expedition. Within a year of the BNSS coming into force in July 2024, Section 223 of the BNSS (“Section 223”) had produced a body of conflicting decisions and a procedural puzzle that courts across the country are navigating differently on any given day. The cause of all this is, on its face, the insertion of a new word and a proviso to the erstwhile statutory text.

Section 200 of the old CrPC required that “a Magistrate taking cognizance of an offence on complaint shall examine upon oath the complainant.” 

Section 223(1) of the BNSS now reads “a Magistrate having jurisdiction while taking cognizance of an offence on complaint shall examine upon oath the complainant.” And then, it carries for the first time: “Provided that no cognizance of an offence shall be taken by the Magistrate without giving the accused an opportunity of being heard.”

Two questions have consequently divided the High Courts. First, does the phrase “while taking cognizance” reposition the complainant’s examination as a step preceding formal cognizance, contrary to the settled position under the CrPC that such examination is post-cognizance? Second, at what point must the accused be heard under the new proviso: immediately upon perusal of the complaint or only after the sworn statements of the complainant and witnesses have been recorded?

This piece attempts a considered analysis of both questions. The competing readings of “while” are then examined, followed by an assessment of why the sequential reading, despite the weight of Supreme Court authority it must contend with, is the more defensible one. The scope of the hearing under the proviso is addressed thereafter, along with considering just what a pending reference in one of the High Courts may resolve. 

The Old Regime and Inherent Ambiguity

Under the CrPC, criminal procedure in private complaint cases followed a sequential model. At the stage of cognizance which was understood as the Magistrate judicially applying mind to the complaint, no hearing was required of either party. A Magistrate simply took notice of the complaint. Examination of the complainant under Section 200 followed, and this examination was understood, on the dominant view, to have been done after cognizance had already been taken i.e., post-cognizance. 

There was, however, an inherent textual ambiguity in the old Section 200 that courts largely papered over (as discussed on the blog), in arriving at this dominant post-cognizance view. 

The opening words, “a Magistrate taking cognizance of an offence on complaint shall examine upon oath the complainant”, are grammatically indeterminate. The lack of clarity over the statute’s use of active voice and present participle (“taking” rather than “having taken”) create a lacuna. It was unclear whether it described an examination that is part of the very act of taking cognizance itself, or if it prescribed a mandatory procedural step consequent to having already taken cognizance.

Left to resolve this friction through construction, the Supreme Court generated two parallel streams of judgments that vacillated on the exact moment cognizance was triggered. One view, articulated in decisions like CREF Finance Ltd. v. Shree Shanthi Homes (P) Ltd., postulated that cognizance is taken at the initial stage when a Magistrate peruses the complaint to ascertain if an offence is disclosed, treating it as a step strictly preceding the examination of witnesses. This was further mapped in S.K. Sinha, Chief Enforcement Officer v. Videocon International Ltd., where the Court held that taking cognizance under Chapter XIV operates as a “sine qua non or condition precedent” before the machinery of Chapter XV, which houses the examination of the complainant, can even commence.

A second, divergent view fused the two events. A four-Judge Bench in Jamuna Singh v. Bhadai Shah laid down the bedrock for this interpretation and noted that the very act of a Magistrate proceeding to examine a complainant under Section 200 of the CrPC “puts it beyond doubt” that cognizance has already been taken. This logic was reaffirmed by a three-Judge Bench in Devarapalli Lakshminarayana Reddy v. V. Narayana Reddy, which clarified that applying one’s mind for purposes of proceeding under Section 200 constitutes taking cognizance. The sequence, as per Nirmaljit Singh Hoon v. State of West Bengal, was that a Magistrate first takes cognizance under Section 200 and then examines the complainant to ascertain if a prima facie case exists to issue process.

Despite this oscillation between treating the examination as the absolute proof of cognizance (per Jamuna Singh) or as a distinct subsequent step (per S.K. Sinha), the dominant consensus, as ultimately reaffirmed by a Constitution Bench in Sarah Mathew v. Institute of Cardio Vascular Diseases, was that Chapter XV of the CrPC functioned unequivocally as a post-cognizance chapter. Therefore, under the old regime, the examination of the complainant followed from, rather than preceded, the taking of cognizance.

Relocating the Examination through the Addition of ‘While’ and the Scope of Hearing

The case for treating the examination as part of the cognizance process rather than as post-cognizance in the new regime runs as follows. Parliament offered no explanatory note on the insertion of “while”, but the word most naturally denotes simultaneity or an ongoing process. “A Magistrate while taking cognizance... shall examine” is most readily read as positioning the examination within the process of taking cognizance and not as something that follows it. 

When read alongside the proviso, which prohibits taking cognizance without first hearing the accused, the implication is that cognizance cannot formally culminate until both the examination and the hearing have occurred. The Madras High Court in Seeman v. Varun Kumar captured this shift and held that “cognizance is no longer a single instantaneous act but a process within which an intermediate point exists at which notice to the proposed accused is envisaged” (¶33). This suggests that the Magistrate is still in the process of arriving at the cognizance decision when examining the complainant.

The Karnataka High Court in Sri Basanagouda R. Patil v. Sri Shivananda S. Patil mapped the sequence to be comprising a complaint filed, complainant examined on oath, notice issued to the accused appending the complaint and sworn statements, hearing, and only then formal cognizance and issuance of process. The Court held that the “question of taking of cognizance would not arise at this juncture”. Instead, this examination generates the substantive material, which refers to the sworn statements, that must be appended to the notice issued to the accused to ensure the hearing is not an “empty formality” (¶10, ¶11). Therefore, notice is issued at that stage, and only after hearing the accused does a Magistrate take formal cognizance and regulate further procedure. So far, the High Courts in Delhi (1, 2), Allahabad, Calcutta (1, 2, 3), Punjab & Haryana, Chhattisgarh, Kerala (1, 2), and Gauhati have adopted the same sequence.

There is also an argument from the proviso’s workability if one attempts to strictly apply the historical Supreme Court standard of cognizance. Per S.K. Sinha, the Court ruled that taking cognizance is the “sine qua non or condition precedent” that must occur prior to the commencement of criminal proceedings. However, the first proviso to Section 223(1) of the BNSS strictly dictates that “no cognizance of an offence shall be taken” without first giving the accused an opportunity of being heard. On this reading, the proviso would be violated the moment the complaint is read. Indian courts have long resisted constructions that produce this kind of structural impossibility or cause procedural paralysis, under the maxim lex non cogit ad impossibilia, the law does not compel what cannot be performed. A provision should not, where another reading is available, be construed so as to make its own compliance impossible from the first moment it applies.

In defining the outer limit, in Seeman, the Madras High Court held that the pre-cognizance hearing permits the accused to make submissions and place defensive material to inform the Magistrate’s evaluation, but absolutely prohibits the cross-examination of witnesses or any full-blown adjudication. However, once again, there is ambiguity regarding the exact utility of this defensive material. Under Section 226 of the BNSS, the dismissal of a complaint must rest exclusively on the sworn statements of the complainant and witnesses (and the result of any Magistrate’s inquiry). Thus, while an accused might successfully highlight a jurisdictional bar or a lack of statutory sanction during the hearing, the Magistrate’s ultimate order of dismissal must be anchored in the inherent weakness of the complainant’s own case.

Furthermore, this threshold hearing also invites unresolved friction regarding Article 20(3) of the Constitution of India. If the accused’s preliminary submissions form part of the record, there is a severe risk that these early defensive assertions could be weaponised against their maker. Until courts clarify whether such pre-cognizance submissions enjoy privilege, this statutory protection risks becoming a self-incrimination trap before the practice of filing detailed written replies becomes entrenched.

The Rita Bakshi Reference and the Doctrinal Dilemma

Perhaps the most detailed exposition between the conflicting positions is found in the Delhi High Court’s March 2026 decision, Dr. Rita Bakshi v. Seema Bajaj & Anr. Dr. Swarana Kanta Sharma, J., while finding herself bound by coordinate-bench decisions (Brand Protectors and Neeti Sharma), noted the structural flaws in the High Courts’ new “sequential reading” and referred the questions of law to a Larger Bench.

While it is to be said that the sequential reading adopted by the High Courts offers a practical exit from the bottleneck created by Section 223(1), it has nonetheless invited severe criticism. This rests on three primary doctrinal pillars: first, the irrelevance of the statutory modifications, second, the impermissibility of retracing judicial steps, and third, the destruction of the speedy trial mandate. Each will be addressed in turn. 

First, it has been advanced that the sequential reading rests on a weak grammatical base and the phrase “having jurisdiction” in Section 223(1) BNSS is dismissed as mere statutory surplusage, since a Magistrate inherently cannot conduct an inquiry without possessing jurisdiction under the statute. Furthermore, the insertion of the word “while” does not alter the fundamental curial act of taking cognizance. Further, as per Supreme Court judgments from R.R. Chari to Jamuna Singh, cognizance is triggered the moment a Magistrate applies their judicial mind for the purpose of proceeding under the chapter. Therefore, on this view, the examination of witnesses cannot be shifted to a pre-cognizance stage.

However, while the addition of “while” is indeed a clumsy legislative change, the High Courts had to improvise to prevent Section 223(1) from becoming self-defeating. If cognizance is treated as occurring the instant a Magistrate reads the complaint, the new proviso would be breached at that very moment. To avoid this impossibility, courts in Brand Protectors and Basanagouda have relied on the word “while” to reconceptualise cognizance as a processual window.

Second, on the view that as per Jamuna Singh, examining the complainant on oath itself shows that cognizance has already been taken. The High Courts’ new “sequential” model (examining witnesses, then issuing notice, then deciding on cognizance) seemingly forces Magistrates to move backwards from post‑cognizance to pre‑cognizance, which they have no power to do since they cannot review or undo cognizance orders. Nonetheless, the counter is that High Courts (like in Basanagouda) avoid this by redefining the timeline and that the examination is treated as pre‑cognizance material-gathering, and formal cognizance occurs only at the end, after hearing the accused. So, when a complaint is dismissed at that stage, the Magistrate is not retracing steps or reviewing an earlier order, but rather simply refusing to take cognizance in the first place.

Third, the sequential reading is criticised for undermining the legislature’s goal of expeditious trials by drawing the accused into a pre‑cognizance “mini‑trial.” An accused could reasonably seek to cross‑examine the complainant, lead defence evidence, or move Section 94 BNSS petitions for document production, spawning revisional and appellate challenges if refused and stalling the case at the threshold. It is undeniable Parliament itself introduced delay through the first proviso to filter out frivolous complaints. The real question is therefore not whether delay exists, but how to make the delayed hearing meaningful. If the strict doctrinal view is followed and the accused is summoned on the bare complaint, the hearing degenerates into an “empty formality” on unsworn allegations. As Basanagouda recognises, a genuine opportunity of being heard requires that notice to the accused enclose both the complaint and sworn statements. The delay is thus an unavoidable consequence of poor drafting. The sequential reading merely ensures that this cost yields real procedural fairness to the accused rather than a hollow ritual.

The delay is the inevitable price extracted by the poor drafting of the BNSS. Regardless, the sequential reading would just ensure that this price buys actual procedural fairness for the accused rather than a hollow ritual.

Conclusion - What the Reference Must (hopefully) Resolve

Section 223 BNSS has plunged cognizance, already the subject of decades of contested jurisprudence, into more complications. This self-inflicted confusion arose owing to the Parliament’s introduction of a substantive safeguard without attending to its interaction with inherited doctrine, and without explaining why. 

A year in, the provision has plainly not made private complaint cases proceed faster. Section 223 BNSS has generated conflicting judicial sequences, inconsistent approaches to the timing of cognizance, and unresolved questions about the scope of the accused’s right to be heard.

The provision also raises broader questions that remain unexplored. As Sekhri observes, it is difficult to discern the principled basis for extending a pre-cognizance hearing to accused persons in private complaints, and in Enforcement Directorate prosecutions such as under the Prevention of Money Laundering Act, 2002 (as per Kushal Kumar Agarwal v. Directorate of Enforcement), while not giving an equivalent safeguard to persons proceeded against through the ordinary FIR route. Whether this differential treatment withstands scrutiny under Article 14 is a constitutional question that may eventually require independent consideration.

If the Rita Bakshi referral reaches the Supreme Court and it endorses the sequential reading, it will accept that some delay is the necessary cost of ensuring an accused is heard on a meaningful evidentiary foundation. If it rejects that reading, it must articulate an alternative construction that gives genuine effect to the proviso without rendering it practically unworkable. Until then, Section 223 will remain less a provision that streamlines criminal procedure and more a tale of caution of how a single poorly drafted legislative amendment can unsettle an entire procedural framework.