Guest Post: Hash-Tagged, Not Verified - What Section 63’s Hash Value Cannot Prove

(This is a guest post by Ratyansh Garg)

On 22 September 2026 the defendant in BPTP Ltd v. Nitin Naresh told the Delhi High Court that a twelve-minute recording of him allegedly demanding money was “in all probabilities a mere AI generated cloned recording of his voice”. Rather than decide the point on the papers, the Court sent the recording to the Central Forensic Science Laboratory “for the purpose of ascertaining the truthfulness and veracity of the recording as also to ascertain the hash value”.

The order treats the truth of the recording and its hash value as two questions. Four months earlier the Supreme Court treated the hash as answering both. Disposing of a challenge to section 63(4) of the Bharatiya Sakshya Adhiniyam 2023 (BSA) in Pune Bar Association v. Union of India, the Court began from the threat that “artificial intelligence and deepfake technology” pose to electronic evidence. It described a hash value as “synonymous with an electronic fingerprint” that and reasoned that the expert’s certification in Part B “provides an additional layer of authenticity”.

This post argues that the Court in Pune Bar asked more of the hash than it can give. A matching hash shows integrity: the file before the court is bit-for-bit the file that was hashed. It says nothing about authenticity, or veracity, whether that file is a genuine record of what it appears to show, which is the relevant inquiry for deepfakes. Conflating the two attributes makes admission costlier for honest litigants and stamps an expert’s signature on records whose veracity no one has verified.

What Section 63(4) BSA now asks for

Section 65B(4) of the Evidence Act required a certificate signed by a person “occupying a responsible official position” in relation to the device. Section 63(4) of the BSA adds three things. The certificate must accompany the record “at each instance where it is being submitted for admission”; it must follow the form in the Schedule; and it must be signed by the person in charge of the device “and an expert”. Part A of the Schedule, filled in by the party, identifies the device, makes sworn declarations about its regular use and proper working, and states the record’s hash value, with a choice of “SHA1”, “SHA256”, “MD5” or “Other”. Part B, filled in by the expert, restates the device particulars and the hash value. As Sulaiman observed, all the expert’s certificate does is “merely replicate the information already provided in Part A”. Beyond the party’s own declarations, neither Part asks anyone to examine whether the record is genuine.

The Court disposed of the petition without issuing notice to the Union, and kept open the question of who may sign Part B. Sekhri has called its reasoning “a lazy recitation of an old logic of tampering”. The present argument is narrower: The order’s premise about what a hash establishes is itself mistaken. 

Integrity is not Authenticity

A cryptographic hash function reduces a file to a fixed-length string. Change a single bit and, the result is “a completely different message digest”. Comparing the hash of a copy with that of the original therefore tells a court whether the copy is identical, and nothing more. The Scientific Working Group on Digital Evidence, a US forensic standards body, explicitly states: “hashing files does not protect against manipulation of evidence before it is hashed”, and “the only way to manipulate the evidence without detection is to do it before it is hashed” (SWGDE, 2019).

Take a shop owner who fabricates CCTV footage of a burglary and hashes the fabricated file. Every later copy will match, and the hash values in Part A and Part B will be accurate. Part A’s declaration that the footage was produced in the ordinary course will be false, but nothing in the certificate tests it, and Part B will certify the fingerprint of a fake. The same holds for a cloned voice note or a doctored screenshot. A hash taken at a trusted moment, such as seizure, does rule out tampering after that moment. Whatever was false before it stays false, now with a matching fingerprint. Mishra made the same point before Pune Bar: section 63 “might act as a safeguard to detect tampering with e-record but it does not vouch for the legitimacy of e-evidence”.

Indian evidence law has long treated tampering as only one of several questions. Ram Singh v. Col Ram Singh set six conditions for tape recordings: the speaker’s voice be identified; accuracy of the recording be proved; “every possibility of tampering with or erasure” be ruled out; and the cassette be sealed and kept in official custody. A hash act as a seal, ruling out alteration after hashing. However, it says nothing about whose voice is on the recording, whether the recording is accurate, or whether it was altered before it was hashed. In Himanshu Chordia v. State of Rajasthan, decided ten weeks after Pune Bar, the Supreme Court asked if photographs sought to be adduced were “genuine pictures” or “doctored using technological advancements”, and observed: “Verification within the meaning of Section 65B of the Evidence Act, 1872… is not the only issue.”

The conflation predates Pune Bar. The Karnataka High Court’s guidelines in Umesh v. State of Karnataka (2022) preferred “cryptographic analysis by hash value method” over spectrographic analysis for assessing “the reliability and authenticity of an electronic document”. Ironically, while spectrography can help identify a speaker; a hash cannot. The Madras High Court in Yuvaraj v. State (2023) directed hash values to be recorded at four stages, reasoning that matching values “would ensure its authenticity”. A Delhi magistrate in State v. Satwant Rao, post Pune Bar, wrote that a hash “assists in verifying the authenticity and integrity of the electronic record”. Careful commentary reads the Schedule the same way, as giving “a more meaningful assurance of the genuineness” of the record (Verma, NLSIR 2025). A trial judge handed a certificate by an “expert” will likely read it as an authenticated forensic finding, which it is not.

A Weaker Fingerprint than the Court assumed.

First, the Schedule offers MD5 and SHA-1. As of now, both have low collision resistance. The Internet Engineering Task Force said in 2011 that MD5 “is no longer acceptable where collision resistance is required” (RFC 6151), and the National Institute of Standards and Technology, under the US Department of Commerce, has announced that SHA-1 should be phased out by 2030 because “collision” attacks “have been used to undermine SHA-1”. These attacks do not let an outsider alter a file that someone else has already hashed; no practical attack of that kind exists. They let a person who controls a file before hashing prepare two different files with the same hash. Forensic practice still accepts MD5 and SHA-1 for integrity checks because “technical and procedural controls during the data collection process” guard against manipulation before hashing (SWGDE, 2019). Part A under has no such controls: the party hashes its own file, and the Schedule lets it hand-pick the weakest algorithm. The practical risk is narrow, since a fabricator who knows that no one examines the content need not bother with collisions at all. But a form whose hash value is treated as an “electronic fingerprint” should not list algorithms that standards bodies have abandoned wherever collision resistance matters.

Second, hashes are brittle in the other direction. A photograph sent on WhatsApp at standard quality is compressed before it is sent, and CCTV footage is often exported in a converted format. The resulting files differ from originals bit-for-bit and so carry different hashes, though no one has tampered with them. A mismatched hash does not prove tampering, just as a matching hash does not prove genuineness. A test that is neither necessary nor sufficient for genuineness is a poor gatekeeper of it.

The Cost at the Gate

None of this would matter much if the certificate were easily accessible, which it is not. Only government bodies can be notified as Examiners of Electronic Evidence under section 79A of the IT Act, and in July 2026 there were reportedly “just 21” of them. Pune Bar suggested that a person with “special skill and expertise in computer science and cyber forensics” may sign Part B if the court is satisfied, but declined to decide the point. Courts have responded unevenly. The Delhi High Court has regularly granted exemptions and extra time for hash and expert certificates (here, here and here). A Delhi civil court held a document inadmissible for want of a section 63 certificate (Devender Kumar v. Gurmeet Singh), and a commercial court held that Pune Bar was inapplicable to commercial suits. (Panasonic Life Solution v. Dinesh Kumar). As Mayank Khichar and Makhija and Sharma argue, the burden falls hardest on private litigants and the defence.

A Wrongly Borrowed Provision, and the Wrong Cure

Section 63 inherits an old problem. As Nariman J noted in Arjun Panditrao Khotkar, section 65B “has its genesis in Section 5 of the Civil Evidence Act 1968 (UK)”. The UK itself repealed that provision with effect from 1997, more than three years before India copied it, and removed its criminal counterpart, section 69 of the Police and Criminal Evidence Act 1984, in 2000. In the same case Ramasubramanian J. called for “a relook at Section 65B”, and the Madras High Court in Yuvaraj said there “could be no better instance of fitting a square peg into a round hole”. The BSA’s response was to add an expert and a hash to the peg. In UK criminal cases, once section 69 went, courts fell back on a common law presumption that, absent evidence to the contrary, “mechanical instruments were in order at the material time”. That presumption too is now under review because of the Post Office scandal, in which sub-postmasters were convicted on data from a system later found to be “not remotely robust” (Hamilton v Post Office). 

What Should Replace it?

The BSA’s validation of electronic evidence can significantly improve with three changes: First, courts must treat a Section 63 certificate as going to integrity alone: as evidence that a copy matches the record that was hashed. Moreover, Courts liberally exempt the production of the s.63(4) certificate. While Arjun Panditrao calls it a condition precedent, it permits a certificate to be produced at any stage before the trial ends. The Delhi High Court also treats a missing certificate as a curable issue, granting time for cure. 

Second, once a party establishes serious doubts to the record’s genuineness, the dispute should go where BPTP sent it: to a forensic examiner under section 39. For deepfakes, the US Advisory Committee on Evidence Rules is discussing a separate draft rule under which the opponent must first produce evidence sufficient to support a finding of fabrication, and the proponent must then show that the item is more likely than not authentic. Such inquiries can use metadata, device logs, detection software, and provenance metadata that India’s amended IT Rules have, since February 2026, required intermediaries to embed in synthetic content made with their tools. But detection tools should serve as evidence, not as certification. Open-source detectors lost about half their performance on deepfakes circulating in 2024 (Deepfake-Eval-2024), and people themselves do little better than chance. Turning a detector’s output into a new admissibility “certificate” would repeat the mistake this post describes.

Third, Parliament should amend Section 63(4) and the Schedule, removing MD5 and SHA-1, requiring the hash to be recorded at the earliest practicable moment: automatically at seizure where systems such as e-Sakshya allow it, and otherwise when the record is first tendered, as the Delhi High Court’s Original Side Rules already require. Then, dropping the filing-stage expert signature reduces costs while losing little, since the expert only checks against tampering while copying anyway.

Guest Post: Protected Witnesses under UAPA and the Bilal Mir Judgment

(This is a guest post by Shashank Sinha)

A bench of the Delhi High Court had the opportunity to consider the interpretation and application of Section 44 of the Unlawful Activities (Prevention) Act 1967 in Bilal Mir v. National Investigation Agency, (Crl. M.C. 5129/2025). The outcome is a peculiar judgement which discusses how important it is that even persons accused of offences under special criminal laws such as the UAPA deserve a trial that is fair and an opportunity to challenge evidence against them by confronting and cross examining witnesses, while at the same time legitimizing declaration of witnesses as protected by way of non-speaking and barely reasoned orders. 

Protected Witnesses in the UAPA

Section 44 of the UAPA allows for certain witnesses for the prosecution to be declared ‘protected witnesses’ at the instance of the witness, the state, or the presiding judge. Declaration of a witness as ‘protected witnesses’ results in their personal information being withheld from the accused throughout the proceedings. Obviously, this has significant ramifications for the accused including their right to bring into question the credibility of the witness and to confront and cross examine the witnesses. Confronting an adversarial witness is one of those rights without which one simply cannot call a trial fair in any sense of the word. It also establishes the credibility of the witness itself as a witness who is able to come out of this confrontation without having any major chinks exposed in their narration reaffirms their truthfulness and integrity in the eyes of the court. 

Since this is a provision that quite heavily curtails rights of the accused that would be considered central to what constitutes a fair trial, it would be interpreted in a very strict [Para 448 in Sahai, J’s opinion in Kartar Singh v. State of Punjab, (1994) 3 SCC 569] manner and would require a thorough judicial application of mind with regards to each witness. 

This much and more has been held by the Supreme Court in its judgement in Mohammad Asarudeen v. Union of India (2025 INSC 746). The Supreme Court held that the Special Court under UAPA is required to individually assess the situation of each witness who is sought to be made a protected witness. The Court also held that a judicial finding should be recorded that there is a danger to the life of the witness should their identity be revealed in open court or to the accused persons prior to and during the course of the trial. It is also essential that this finding be not mechanical and shows proper application of mind. Thus, the Court laid down clearly that any order declaring any witness for the prosecution as ‘protected’ under Section 44 UAPA cannot be vague, omnibus and non-speaking given the massive implications such an order has on the right of the accused to bring their character, integrity and their substantive testimony into doubt during the trial. 

Diluting Mohammad Asarudeen and ignoring it altogether

The first part of Bilal Mir correctly discusses the law on protected witnesses under UAPA and states, in slightly more detail, what has been stated in the preceding paragraphs. However, it is in applying this standard where the Court goes wrong. The Court dilutes the standard of judicial application of mind required of a Special Court under UAPA in making such an order. Where Mohammad Asarudeen requires it to reach a specific judicial satisfaction as to there being a danger to the life of the witness before passing an order under Section 44, the Court here reduced the standard to mere ‘probability’ of danger which is a substantially reduced standard. This very holding puts Bilal Mir in the teeth of Supreme Court precedent. 

Secondly, there is also a point where the Court concedes that the Special Court’s order which was passed ex parte and without referring specifically to witnesses who were sought to be made protected witnesses by the prosecution might not be in consonance with the judgement of the Supreme Court in Mohammad Asarudeen (Paras 10 and 79). Yet the High Court’s decision as to the validity of the impugned order does not turn on this. The High Court instead points to the fact that the order was not immediately challenged even when applicants had also filed an application for copies of unredacted statements of the protected witnesses in the wake of the impugned order. Thereafter, the High Court inexplicably reasoned that since applicants did not raise any particular grievance before the Special Court while it was considering the application under Section 44, they must be precluded from challenging the same before the High Court (Para 80). Needless to say, ex parte proceedings are not invitations to dispense with the requirements of judicial application of mind and pass mechanical orders. [State Bank of Patiala v. Vinesh Kumar Bhasin, (2010) 4 SCC 368 - Para 20]. Regardless, even the Court’s assertion that applicants did not challenge either during the Section 44 proceedings or after the order had been passed is not true given that the application for unredacted statements of the protected witnesses was not entertained by the Special Court on the ground that it ‘amounted to a review of its earlier order’. This could very well have been construed as a challenge as the Court itself records. 

Finally, the most peculiar part. The High Court concedes that the Impugned Order did not follow binding Supreme Court precedent. Yet, later on, it tried to justify the validity of the Impugned Order stating that the Special Court had considered the entire chargesheet before passing orders under Section 44. Where this was apparent from the impugned order, and whether consideration of the chargesheet alone is adequate to reach the ‘satisfaction’ that Mohammad Asarudeen demands of the Special Court passing an order declaring a witness as ‘protected’, the Court does not say. 

Conclusion

This might not be the last time that mechanical orders declaring witnesses as ‘protected’ in UAPA prosecutions and binding Supreme Court precedent is ignored. Considering what is at stake, it simply cannot become the norm that Special Courts under the UAPA can pass unreasoned mechanical orders to declare witnesses as ‘protected’ given the consequences for the accused during the trial and the harsh punishments that follow if the state is able to secure a conviction.

Bilal Mir may be cited as an instance of watering down extremely essential safeguards for the accused in prosecutions where the balance is already extremely tilted in favour of the state. However, considering how the Court placed such importance on the failure of the applicants to challenge the order, one can argue that the observations in Para 83 ought not to be treated as its ratio. 

(Thanks to Aadi Behle for his comments and insights).

Guest Post: Per Incuriam and the Appellate Tribunal's Reading of Section 8(3) PMLA

(This is a guest post by Riddhi Alok Puranik)

A May 2026 order of the Appellate Tribunal constituted under the PMLA 2002 (“ATPMLA”), disposing of four appeals in the NSEL investigation, does something I argue that a Tribunal is not permitted to do. Confronted with a Delhi High Court Division Bench judgment going against how the Directorate of Enforcement read the PMLA, the Tribunal neither followed that authority nor distinguished it on the facts. It held the judgment — Directorate of Enforcement v. Rajesh Kumar Agarwal — to be per incuriam and declined to follow it.

The underlying dispute is narrow. Section 17 of the PMLA allows an authorised officer to search and seize or freeze property. Section 20 allows an authorised officer, on fresh recorded reasons, to retain that property for up to 180 days. Section 8(3) allows the Adjudicating Authority (also constituted under the PMLA) to confirm a retention. Rajesh Kumar Agarwal reads these as sequential i.e. without a Section 20 order the retention is void, and there is nothing for section 8(3) to confirm. The Directorate's case, which the Tribunal accepted, is that Section 8(3) carries an independent power to order retention on a Section 17(4) PMLA application, so that Section 20 may be bypassed altogether.

That interpretive question is contested and is now before the Supreme Court. The Tribunal was entitled to think the Delhi High Court wrong about it. What it could not have done, in my respectful submission, was to declare a Division Bench judgment as per incuriam. This is for two independent reasons: the doctrine does not fit a judgment that construed the very provision said to have been overlooked, and a Tribunal subject to a High Court’s superintendence cannot invoke this rule to avoid a judgment it disagrees with. This piece takes both objections in turn, then sets out why the ED’s reading of section 8(3) fails on the ordinary tools of construction, and finally what is at stake institutionally if the Tribunal’s approach stands.

The Interpretive Question

Three provisions of PMLA govern what happens to property after an ED search:

• Section 17(1) empowers authorised officers to search and seize, or where seizure is impracticable, freeze property, on a recorded reason to believe. Section 17(2) requires the officer to forward those reasons and material to the Adjudicating Authority immediately. Section 17(4), inserted by the 2013 amendments, requires the officer, within thirty days, to file an application before the Authority "requesting for retention of such record or property seized....or for continuation of the order of freezing."
• Section 20 is titled "Retention of property". Section 20(1) provides that where property has been seized or frozen under Section 17, an officer authorised by the Director who has recorded fresh reasons to believe that the property "is required to be retained for the purposes of adjudication under Section 8" may retain or continue the freeze for up to 180 days. Section 20(2) requires that retention order and accompanying material to be forwarded to the Authority. Section 20(3) mandates return of property on expiry of 180 days unless the Authority permits continuation. Section 20(4) requires the Authority, before permitting such continuation, to satisfy itself that the property is prima facie involved in money-laundering and required for adjudication.
• Section 8(3) provides that where the Adjudicating Authority decides under Section 8(2) that property is involved in money-laundering, it shall, "by an order in writing, confirm the attachment of the property made under sub-section (1) of section 5 or retention of property or record seized or frozen under section 17 or section 18" whereupon the attachment or retention may continue during investigation for up to 365 days, or the pendency of proceedings.

The dispute here is about the "or" in Section 8(3) PMLA and the operative verb. Does "confirm" distribute across both objects, so that the Authority confirms either a Section 5 PMLA attachment or a Section 17/18 PMLA retention that must first exist under Section 20? Or, does "or" introduce a separate, verb-free power for the Authority to directly order retention on a section 17(4) application, bypassing section 20 entirely?

What the Delhi High Court Held, and What has Followed

In Rajesh Kumar Agarwal, the Delhi High Court read the PMLA scheme as sequential, where Section 17 supplies the seizure/freeze power, Section 20 supplies the executive-layer retention authority for up to 180 days, and Section 8(3) supplies quasi-judicial confirmation for any continuation beyond that. Section 17(4) PMLA is the procedural mechanism through which the officer brings the matter before the Authority and is not an independent retention mechanism [Rajesh Kumar Agarwal (41)–(60)]. Retention not founded on a Section 20 order is void ab initio, incapable of being cured by later confirmation [Anirudh Pratap Agarwal (88)].

This decision is, for now, settled law in the Delhi High Court. Anirudh Pratap Agarwal v. Enforcement Directorate called Section 20 "the crucial link" between seizure and adjudication [Anirudh Pratap Agarwal (34)]. Directorate of Enforcement v. Poonam Malik followed. Ravi Aggarwal v. Deputy Director, Enforcement Directorate treated Rajesh Kumar Agarwal as binding precedent and held that section 8(3) "only governs confirmation."

Why the Per Incuriam Label Fails

The doctrine does not fit on its own terms

A decision is per incuriam when rendered in ignorance or forgetfulness of an inconsistent statutory provision or a binding authority. The AT-PMLA itself quoted Morelle v. Wakeling and Municipal Corporation of Delhi v. Gurnam Kaur to this effect [M/s N K Proteins (58), (60)], correctly stating that decisions given "in ignorance of the terms of a statute" may be treated as per incuriam. But Rajesh Kumar Agarwal was not rendered in ignorance of section 8(3): it construed that provision exhaustively alongside Sections 17 and 20. 

The ATPMLA held that Rajesh Kumar Agarwal is in conflict with the Supreme Court's judgment in Vijay Madanlal Choudhary v. Union of India. This claim has a structural problem though which the ATPMLA did not confront: Vijay Madanlal Choudhary was a constitutional challenge to the PMLA's enforcement architecture, not a case where the precise sequencing of Sections 20 and 8(3) was in issue. Paragraph 84 of that judgment describes the Section 17(2) and 17(4) safeguards in general terms, noting that the Authority "passes an order of retention" following a show-cause process, as part of the Supreme Court's analysis that the overall scheme contains inbuilt safeguards rendering it constitutionally valid.

It does not decide, because it was not asked to decide, whether a Section 20(1) order is a mandatory precondition to the Authority's section 8(3) power. A constitutional imprimatur on the overall architecture is not a specific holding on a sequencing question within that architecture. A court that subsequently decides that sequencing question does not thereby "ignore" the constitutional approval judgment. The Tribunal treated these as interchangeable when they are not.

Nor does the supposed conflict with Vijay Madanlal Choudhary sustain the per incuriam label even on its own terms. There is a fundamental difference between saying that a judgment is inconsistent with, or departs from, Paragraph 84 and saying that it is per incuriam because paragraph 84 was overlooked. Per incuriam requires ignorance of a binding, applicable ratio; it does not require agreement with every prior holding on a general scheme. The ATPMLA conflated disagreement with a precedent and holding that the precedent was per incuriam, which are entirely different propositions.

A lower adjudicatory body may disagree with a High Court's reasoning, but it lacks the institutional authority to treat that precedent as non-binding on that basis alone.

A Tribunal cannot in any case apply the label to a High Court

Even if the doctrine technically applied, the ATPMLA cannot invoke it against a judgment of the Delhi High Court. This is directly settled. In Union of India v. Chand Singh, the Delhi High Court held that "it is not open to the Tribunal to characterise a judgment of a High Court as per incuriam." [Chand Singh (7)]. One narrow latitude was recognised: where a tribunal has before it Supreme Court judgments that state the law differently from a High Court, it may follow the Supreme Court in preference. That is not a licence to call the High Court wrong on the statute; it is a duty, in cases of genuine conflict, to follow the higher court. That latitude is unavailable here because the claimed conflict with Vijay Madanlal Choudhary does not exist. 

What we have is a statutory tribunal substituting its own reading of section 8(3) for that of the High Court. The constitutional architecture forecloses this. The Supreme Court in East India Commercial Co. v. Collector of Customs held that a tribunal subject to a High Court's superintendence cannot ignore the law that court has declared and proceed in violation of it; if it could, the entire logic of supervisory jurisdiction collapses [East India Commercial Co Ltd (14), (31)]. L. Chandra Kumar v. Union of India placed High Court superintendence over tribunals within the basic structure of the Constitution [L. Chandra Kumar (79), (92)].

Why the ED's Interpretation Cannot Hold: A Convergence of Methods

The foundational error does not depend on the underlying interpretation being wrong. But it is worth examining why multiple orthodox techniques converge against the ED's position, because the Tribunal treated the grammar point as sufficient without considering what it does to the rest of the Act.

Grammar and the distribution of "confirm"

The subject-verb-object structure of section 8(3) is: "he shall, by an order in writing, confirm the attachment... or retention... under section 17 or section 18". "Confirm" is the operative verb, governing a single object list connected by "or." On the ED's reading, "or retention" introduces a power to originate retention, i.e. a power for which there is no verb in the sentence and no source in Section 8's architecture.

The Tribunal read "or" as disjunctive and derived from it a new, separate power. But "or" being disjunctive establishes only that there are two distinct objects; it does not supply a second, different verb [M/s N K Proteins (37)]. Filling that gap requires reading words into Section 8(3) that Parliament did not write, which is precisely what the Supreme Court in Vijay Madanlal Choudhary itself said constitutes rewriting the statute [Vijay Madanlal Choudhary [33)].

The Section 5 PMLA Structural Argument

The most telling textual point comes from a direct comparison with Section 5 of the PMLA. Section 5(1) reads: "he may, by order in writing, provisionally attach such property for a period not exceeding one hundred and eighty days from the date of the order". This language is self-contained. It grants the Director an executive power of attachment directly, sets its own duration, and requires no separate provision to authorise retention. Section 8(3) then "confirms the attachment... made under sub-section (1) of section 5" and it confirms a prior executive order that Section 5 itself generates.

Section 17’s drafting rationale clearly departs from section 5. It could have, in order for it to be construed as a similar self-containing code, included language providing that the seizing officer may, by order in writing, retain such property for a period not exceeding 180 days. It does not thus, replicate the logic of section 5. Section 17 authorises officers to seize or freeze and contains no self-standing retention authority whatsoever. The officer who seizes under Section 17(1) has no power under section 17 to retain. Section 17(4)'s language confirms this: the officer files "an application requesting for retention", the language of petition to another body, not the language of independent executive order.

The retention authority for Section 17 PMLA property was placed, deliberately and separately, in Section 20. Section 20(1) mirrors section 5(1) in structure and provides that property "may, if seized, be retained... for a period not exceeding one hundred and eighty days". That is the executive retention order. Section 8(3) then confirms it, in the same manner that it confirms the section 5 attachment in the attachment track.

The expressio unius (Latin: expression of one thing is exclusion of another) inference is straightforward.  Parliament clearly enacted two enforcement tracks in the same statute. Under the attachment framework in Section 5, the authority to act and the authority to retain the property for 180 days are contained within a single provision. By contrast, under the search-and-seizure framework, Parliament distributed those functions across Sections 17 and 20, with Section 17 authorising the initial seizure and Section 20 governing subsequent retention. This structural choice reflects deliberate legislative design, where both tracks converge at section 8(3) for quasi-judicial confirmation but both require a prior executive order.

The ED’s argument effectively treats a seizure under Section 17 PMLA as the functional equivalent of an attachment under Section 5 for the purposes of Section 8(3). On this view, the Adjudicating Authority may confirm the seizure itself, even in the absence of a prior order under Section 20. Such an interpretation collapses the distinction that the statute deliberately maintains between the attachment and seizure mechanisms. If a seizure under Section 17 were by itself sufficient to attract Section 8(3), Section 20 would cease to perform any independent role within the search-and-seizure framework. The result would be to render Section 20 redundant, which is a consequence that the statutory scheme does not support.

Differential predicates

Section 20(1) PMLA obliges the authorised officer to record reasons showing that the property must be retained for the purposes of adjudication under Section 8. Section 8(2), which serves as the basis for any order under Section 8(3), requires the Adjudicating Authority to make a substantive finding that the property is involved in money laundering. These are materially different standards operating at different stages in the procedural machinery. The Section 20(1) determination is only a preliminary, executive-level judgement about investigative need while the Section 8(2) determination is a quasi-judicial finding on merits.

If the Section 17(4) PMLA application goes directly to the Adjudicating Authority without any section 20 order, the Authority is asked to make a full-blown section 8(2) money-laundering determination at the earliest possible post-seizure stage, before investigation is substantially complete, basis only the material forwarded under section 17(2). Parliament's decision to give the ED 180 days under section 20 exists precisely because that determination cannot responsibly be made immediately. Collapsing these tiers forces the Adjudicating Authority into an adjudication on merits at a stage for which the statutory scheme provides no adequate procedural preparation.

The Convention architecture

The PMLA's Statement of Objects and Reasons, reproduced in the Solicitor General's Note II, names the United Nations Convention Against Illicit Traffic in Narcotic Drugs and Psychotropic Substances 1988 as one of the international instruments calling for prevention of laundering of drug proceeds and confiscation of proceeds derived from such offences. Chapter VA of the Narcotic Drug and Psychotropic Substances Act, 1985 (“NDPS Act”) was enacted to implement the same Convention's Article 5 on confiscation, and its seizure-confirmation architecture under Section 68F is a two-step structure where the investigating officer seizes or freezes under section 68F(1), and the quasi-judicial competent authority confirms within 30 days under section 68F(2), failing which property is returned.

That confirmation power is precisely a power of confirmation, not an original power of seizure. PMLA was enacted to address the money laundering which the NDPS regime had left uncovered, extending the Convention's framework to the full range of scheduled offences. If the ED's reading of Section 8(3) as an original retention power were correct, that expanded statute would provide fewer executive-layer accountability steps in its search-and-seizure track than the narrower NDPS regime it was designed to go beyond. That outcome can’t be reconciled with the Centre's own framing of PMLA as giving effect to India's Convention obligations.

The Institutional Stakes

The Tribunal’s interpretive disagreement with the Delhi High Court can very well be legitimate. The question remains contested, and with both Rajesh Kumar Agarwal and Anirudh Pratap currently pending before the Supreme Court, awaiting the Apex’s resolution on interpretation. The difficulty lies not in the Tribunal’s disagreement, but in the institutional consequences it drew from it.

The ATPMLA sits in a legal order where it is subject to Delhi High Court jurisdiction under Articles 226 and 227 of the Constitution and, hears appeals that travel to the High Court under Section 42 of PMLA, and has before it a consistent body of Division Bench authority directly on point. If it disagrees with that authority, the remedies open to it are narrow and well-defined, which are to follow the High Court while recording the difficulty and distinguish on facts where facts genuinely differ. Declaring a Division Bench judgment per incuriam is not one of those remedies.

The practical consequences of the Tribunal's approach, left unchallenged, are that Section 20 protections become effective only for those who can afford to litigate back to the High Court under Section 42. The Adjudicating authority cites the ATPMLA and thus it has immunised itself from High Court correction. Property seized without a Section 20 PMLA order stays seized, and Article 300A operates at the Tribunal's discretion rather than as a function of law. That is not a procedural quibble. It is the difference between a safeguard that functions and one that exists only on paper.

Guest Post: Indiscriminate by Design - Case against Pellet Guns

(This is a guest post by Rudraksh Lakra)

Pellet guns entered the conversation around Indian protest policing in July, when the Rapid Action Force used them against demonstrators at the Cockroach Janta Party (“CJP”) march in Delhi. This is not new territory; security forces have deployed pellet-firing shotguns against protesters in Jammu and Kashmir since the early 2010s.

While hearing petitions concerning the violence during the CJP protest, the Supreme Court observed that pellet guns may be used in exceptional circumstances. This approach mirrors the Court's response in 2017 when it considered the excessive use of pellet guns in Jammu and Kashmir. The Court observed that pellet gun deployment ought to require proper application of mind by the authorities.        

I find this approach deeply unsatisfactory. In 2017, the Court did not seriously examine whether pellet guns were themselves constitutionally permissible. In 2026, even at the stage of interim relief, the Court observed that pellet guns may be used in exceptional circumstances, nudging against a constitutional challenge to the weapon itself. In both instances, the Court chose to regulate the manner in which pellet guns may be deployed rather than asking whether they are compatible with the constitutional guarantees of life and peaceful assembly.

In contrast, I argue that the use of pellet guns for policing protests fails the proportionality test because they are inherently indiscriminate and excessive by design. Unlike conventional ammunition that discharges a single projectile, pellet cartridges release dozens or even hundreds of metal pellets in India that spread over a widening area immediately after being fired. Once fired, each pellet travels independently, making it impossible to confine the force to a particular individual. The foreseeable area of impact extends beyond the intended target, exposing nearby protesters, journalists, bystanders, and even persons attempting to leave the assembly to serious injury. 

The resulting injuries are determined more by the mechanics of the ammunition than the decision to pull the trigger. This matters because it moves the inquiry away from the conduct of the individual officer. An officer who exercises restraint, aims carefully, and fires in line with official instructions still cannot control where each pellet lands once it leaves the barrel. The risk of striking unintended persons is built into the weapon rather than caused by poor implementation. Procedural safeguards can govern the decision to fire, but they have no bearing on the trajectory of a pellet after it has been discharged. 

The severity of injuries caused by pellet guns further demonstrates why they are unsuitable for policing assemblies. Metal pellets penetrate soft tissue and frequently strike highly vulnerable parts of the body, particularly the face and eyes. Victims can suffer permanent blindness, retinal detachment, ruptured eyeballs, facial fractures, internal bleeding, nerve damage, lung injuries, and abdominal trauma (see here, and here). In Jammu and Kashmir, between 2015 and 2017, pellet guns were responsible for the deaths of at least 17 people. The Jammu and Kashmir Chief Minister, Mehbooba Mufti, informed the Legislative Assembly that between July 2016 and February 2017 alone, 6,221 persons sustained pellet injuries, including 782 eye injuries. More than 1,000 Kashmiris have suffered partial or complete loss of vision since pellet guns were introduced in 2010.

This is also why leading human rights organisations have consistently concluded that pellet guns are unsuitable for policing assemblies. Physicians for Human Rights and the International Network of Civil Liberties Organizations concluded that pellet rounds disperse multiple projectiles across a broad area, making accurate targeting impossible. For this reason, they recommend that kinetic impact projectiles firing multiple pellets should be prohibited during protests. The Omega Research Foundation similarly explains that the spread pattern of pellets is inherently inaccurate and that the pellets can lodge in multiple parts of the body, making medical treatment significantly more difficult.

The United Nations guidance on less lethal weapons in law enforcement notes that weapons firing multiple projectiles generally cannot comply with the principles of necessity and proportionality, and recommends banning metal pellets, fired from shotguns like in India.  Human Rights Watch also calls upon India to prohibit the use of metal pellet shotguns for crowd control. This inability to distinguish between individuals, and its capacity to inflict serious physical injuries, renders it incompatible with international human rights law. 

Therefore, a constitutional framework founded upon necessity and proportionality cannot accommodate a weapon whose ordinary operation creates a foreseeable risk of striking persons who were never intended to be hit. 

The same conclusion follows from domestic law governing the use of force. Sections 148 to 150 of the Bharatiya Nagarik Suraksha Sanhita, 2023 (“BNSS”), corresponding to Sections 129 to 131 of the Code of Criminal Procedure (“CrPC”), embody a principle of graduated force while dispersing unlawful assemblies. Section 149(3) of the BNSS, corresponding to Section 130(3) of the CrPC, expressly requires that the armed forces use "as little force, and do as little injury to person and property, as may be consistent with dispersing the assembly." In Anita Thakur v. Union of India (2016), the Supreme Court held that force used to disperse an unlawful assembly must remain reasonable and confined to what is necessary to restore order. In Ramlila Maidan Incident v. Home Secretary (2012), the Court recognised that excessive force by the police, including disproportionate use of tear gas, and noted that these actions had “an element of criminality”. A weapon that is inherently indiscriminate, and excessive sits uneasily within a statutory scheme built around a graded, minimal, reasonable, and proportionate response.

For these reasons, I believe the Supreme Court's protocol and safety-based approach to pellet guns addresses a symptom rather than the underlying problem. A weapon whose ordinary operation predictably disperses uncontrollable projectiles cannot be rendered proportionate through better training, stricter authorisation or revised operational guidelines. The documented experience of Kashmir demonstrates that the harms associated with pellet guns are systemic rather than accidental. The constitutional guarantees of life and assembly therefore require a categorical prohibition on the use of pellet guns in protest policing.