Law & Policy
The Notice Before the Arrest: What Section 35(3) Leaves for the Remand Order to Fix
31 August 2026 · 8 min read
Section 35(3) of the Bharatiya Nagarik Suraksha Sanhita, 2023 requires a police officer, in every case where arrest is not required under sub-section (1), to issue a notice directing the person to appear rather than arresting him. The person who complies, and continues to comply, is not to be arrested for that offence unless the officer records reasons for thinking arrest is now called for. This is not a new rule. It is Section 41A of the Code of Criminal Procedure, 1973, carried forward with the operative word — "shall" — unchanged from the form that provision took after the 2010 amendment. And the reason that word matters at all was settled, for most practical purposes, not by either Parliament but by a judgment: Arnesh Kumar v. State of Bihar, (2014) 8 SCC 273.
Three law-making acts, spread across thirteen years, have now produced the same command: notice instead of arrest, for offences carrying up to seven years, unless the officer has and records a reason to do otherwise. It is worth asking what each of the three actually contributed, given how little the text itself moved.
The command and its addressee
Section 41(1)(b) CrPC — reproduced with the same structure in Section 35(1) BNSS — does not make arrest automatic on a credible complaint of a cognizable offence punishable up to seven years. It requires two things: a reason to believe the person committed the offence, and a separate satisfaction that arrest is necessary for one of a specified set of purposes — to prevent the person from committing a further offence, for proper investigation, to prevent him from tampering with or destroying evidence, to prevent him from inducing or threatening witnesses, or to prevent him from absconding from the process of law, and in the last case a written record of the reason. Section 41A then addresses the residual case: where none of those purposes is made out, so that arrest is not required, the officer shall issue a notice instead.
Read together, the two provisions are not an exception carved into a general power of arrest. They are the reverse: arrest for this category of offence is conditioned on an affirmative, purpose-specific finding, and notice is the default position that operates whenever that finding is not made. What is unusual, for a command of this kind, is who it is addressed to and where it operates. It does not run to a magistrate, who sits in a courtroom with a record being kept. It does not run to the accused, who could invoke it. It runs to a police officer, exercised alone, in the field, at the moment a complaint or piece of information is received — a moment at which, by definition, no other institutional actor is present to observe whether the officer in fact turned his mind to the two questions the section poses, or simply arrested because arrest was the easier and more familiar response.
A duty exercised at a moment with no witness is not thereby meaningless — the officer's own case diary is supposed to record the satisfaction reached, and that diary can later be examined. But it means the duty has, structurally, no contemporaneous forum. Nobody is positioned to ask, as the arrest is happening, whether the notice should have issued instead. Whatever enforcement the section gets, it gets afterward.
What Arnesh Kumar actually added
By 2014, when Arnesh Kumar was decided, Section 41A had already existed as a mandatory provision for four years. The judgment is frequently described as having introduced the notice-before-arrest rule; it did not — the rule was on the statute book, in that form, before the case was filed. What the judgment did was address the fact that the rule was not, in practice, being followed, and it did so by fixing on a different actor and a different moment: not the police officer at the point of arrest, but the magistrate at the point of authorising further detention.
The direction was that a magistrate, before authorising detention, must be satisfied — and must record that satisfaction — that the conditions for arrest under Section 41 were in fact met, rather than accepting the fact of arrest as sufficient in itself. That is a check placed at the first production of the accused, which under Article 22(2) of the Constitution and the corresponding procedural provision — Section 167 CrPC at the time, Section 187 BNSS now — must occur within twenty-four hours of arrest. It does not require the accused to do anything. It is not an application he has to file, a ground he has to plead, or a delay he has to survive before the safeguard becomes available to him. It happens, if the direction is followed, by operation of the same order that puts him into custody in the first place.
That is the structural contribution worth isolating. Arnesh Kumar did not create the underlying duty; the duty was already there in the text of Section 41A. What it did was move the earliest point at which a court could test compliance with that duty as far forward as the procedural calendar allows — to the remand order, rather than to whatever later stage the accused might get around to raising it.
The stage after that
Satender Kumar Antil v. Central Bureau of Investigation, (2022) 10 SCC 51 went a step further, and did so by changing what a court is required to do with non-compliance once found, rather than when it is found. The Court held that failure to comply with Sections 41 and 41A entitles the accused to bail — not merely to an argument that the arrest was irregular, but to the relief itself, treating the omission as sufficient rather than requiring the accused to separately establish prejudice. Investigating agencies, the Court said, are duty-bound to comply with the mandate of the two sections. This operates on a different footing from the remand-stage check Arnesh Kumar addressed: it is available at whatever stage the accused is able to bring it — including well after the twenty-four-hour window has closed and detention has continued — but it converts the finding of non-compliance into an outcome rather than leaving it as one factor among others for a court exercising bail discretion to weigh.
Between the two, the position by the time Section 35 BNSS was drafted was this: the duty to issue notice ran to the police officer; a check on whether the duty had been honoured had been located at the remand order, occurring automatically within twenty-four hours; and a further, independent consequence — entitlement to bail — attached at whatever later stage the question was raised and made out. None of that architecture required new statutory text. It had all been built by reading the existing Section 41A alongside Article 22(2) and the general law of bail.
What the recodification carried forward, and what it did not touch
Section 35 BNSS restates the arrest conditions and the notice provision in the same structure, with sub-section (3) preserving the "shall issue a notice" language and the same qualification — that a person who complies is not to be arrested unless the officer, for reasons to be recorded, forms the view that arrest is now warranted. Section 187 BNSS preserves the twenty-four-hour production requirement that gives the remand order its role as the earliest checkpoint. What the recodification did not do, because no recodification of this particular provision could do it, is give the underlying duty a forum at the moment it is actually exercised. The notice-or-arrest decision is still made by one officer, alone, before any court is seized of the matter. That has not changed across 2010, 2014, or 2023, because it could not — the arrest is definitionally the event that first brings a court into the picture, and a duty that operates only up to the point of arrest cannot, by its own logic, be reviewed by a court before the arrest has already happened or not happened.
What each of the three episodes changed was something more modest, and worth stating in those terms rather than as an incremental strengthening of "the law on arrest" in general. The 2010 amendment made the notice obligatory rather than discretionary in the officer's own hands. The 2014 judgment did not add to that obligation; it fixed the earliest moment at which a different actor — a magistrate — could test whether it had been honoured, and located that moment at the point the accused is first produced, so that the check does not depend on the accused's own initiative. The 2022 judgment fixed what follows once non-compliance is found, wherever in the process it is found. The 2023 recodification carried the resulting structure forward without alteration.
None of this closes the gap between the moment the section is meant to govern and the moment a court can look at it. That gap is not a drafting defect capable of being amended away; it follows from what an arrest is — the point at which judicial supervision of the police first becomes possible, not a point at which it was already available. The three enactments did not shorten the distance between the officer's decision and a court's view of it to zero, and no fourth one will. What they did, each in a different way, was shorten the distance between the arrest and the first opportunity for a court to ask whether it should have happened at all.
Authored by Eshan Kumar Gupta, Advocate, Hon'ble Allahabad High Court, Lucknow Bench.