GAAR Part 3: judicial doctrine, Circular 7 of 2017 and defending a reference

In one line. The statute is not the whole of the anti-avoidance law a GAAR file has to answer: there is a judicial doctrine that ran before Chapter X-A and still runs beside it, a Board circular that will be produced across the table, and a tax audit clause that has already asked the question once, in a document signed before any notice issued.

This is Part 3 of three. Part 1 sets out the substantive test under section 96 and the two gates in Rule 10U. Part 2 takes up section 99 and the three crore aggregate, the section 98 consequences, the section 100 question, the section 144BA machinery, the remedy, and what the three Telangana High Court decisions actually decide. This Part is the law that sits beside the Chapter, and the practice: the judicial anti-avoidance doctrine, what Walfort decided about dividend stripping and what is wrongly attributed to it, Circular No. 7 of 2017, the Department’s case at its highest, the Form 3CD clause 30C exposure, how a reference is answered, how a transaction is built so that the question does not arise, and the full concordance with the Income-tax Act, 2025.

The facts in the background are those of M/s. EKGE Retail LLP v. Assistant Commissioner of Income-Tax, W.P. No. 21210 of 2022, in which the Telangana High Court quashed Approving Panel directions on exchange-traded bonus stripping on 13 October 2026. They are set out in full in Part 2.

Judicial anti-avoidance alongside the statute

GAAR did not arrive on empty ground, and the older doctrine still matters, both because it governs pre-2018 years and because the Revenue argues that it survives alongside the statute.

McDowell, and the attribution that is usually got wrong. McDowell & Co. Ltd. v. Commercial Tax Officer, Civil Appeal No. 570 of 1983, was decided on 17 April 1985 by a Constitution Bench of Chandrachud CJ, Chinnappa Reddy, Desai, Ranganath Misra and Venkataramiah JJ, and is reported at (1985) 3 SCC 230 and (1985) 154 ITR 148. The passage everybody quotes is:

“Tax planning may be legitimate provided it is within the framework of law. Colourable devices cannot be part of tax planning and it is wrong to encourage or entertain the belief that it is honourable to avoid the payment of tax by resorting to dubious methods. It is the obligation of every citizen to pay the taxes honestly without resorting to subterfuges.”

That passage is Ranganath Misra J’s, for the majority. It is routinely attributed to Chinnappa Reddy J, and the word “colourable” does not appear anywhere in his separate opinion. What is Chinnappa Reddy J’s is the far wider proposition that the time had come to depart from the Westminster principle and that it had been given “a decent burial”, and it is that proposition, not the colourable devices passage, which the Supreme Court later declined to adopt.

Azadi Bachao Andolan drew the line. In Union of India v. Azadi Bachao Andolan, Civil Appeals Nos. 8161-8162 and 8163-8164 of 2003, decided on 7 October 2003 by Ruma Pal and Srikrishna JJ with the judgment delivered by Srikrishna J, reported at (2004) 10 SCC 1 and (2003) 263 ITR 706, the Court upheld Circular No. 789 and declined to treat treaty shopping as unlawful in the absence of a limitation of benefits clause. On McDowell it was explicit. It introduced the colourable devices passage as the words of Ranganath Misra J “speaking for the majority”, described that opinion as “a far cry from the view of Chinnappa Reddy, J”, and concluded:

“We are afraid that we are unable to read or comprehend the majority judgment in McDowell as having endorsed this extreme view of Chinnappa Reddy, J”

It added a caution that remains useful in any GAAR hearing:

“These words are not intended to be used as magic mantras or catchall phrases to defeat or nullify the effect of a legal situation.”

One point of precision for anyone quoting the taxpayer’s right to arrange his affairs: Azadi Bachao does not state it in its own voice, and it is commonly misattributed. The proposition that the subject may arrange his affairs so as not to attract tax is quoted there from Lord Sumner in Fisher’s Executors; the Duke of Westminster passages reproduced are Lord Atkin’s, on the subject’s legal right to dispose of his capital and income so as to attract the least tax, and Lord Tomlin’s, on documents used as a cloak to conceal a different transaction. Lord Russell of Killowen is mentioned only at one remove, as having been quoted with approval in an earlier Constitution Bench decision. The Court also quotes with approval the Gujarat High Court in Banyan and Berry. Its own nearest holding is that an act which is otherwise valid in law cannot be treated as non-est merely on the basis of some underlying motive. The practical answer to all of this is now EKGE itself, which states the proposition in its own voice and does so inside Chapter X-A rather than outside it.

Vodafone settled the method. Vodafone International Holdings B.V. v. Union of India, Civil Appeal No. 733 of 2012, was decided on 20 January 2012 by Kapadia CJI, Radhakrishnan and Swatanter Kumar JJ, the main judgment delivered by Kapadia CJI for himself and Swatanter Kumar J with a separate concurring judgment by Radhakrishnan J, reported at (2012) 6 SCC 613. At paragraph 71 the Court held that section 9(1)(i) “is not a ‘look through’ provision” and that providing for look-through “is a matter of policy”. At paragraph 73 it held that “In this case we need to apply the ‘look at’ test”, that “the task of the Revenue is to ascertain the legal nature of the transaction” and that “it has to look at the entire transaction holistically and not to adopt a dissecting approach”. At paragraph 64 it held that “it cannot be said that all tax planning is illegal/illegitimate/impermissible”, that “colourable device cannot be a part of tax planning”, and that “there is no conflict between McDowell and Azadi Bachao”.

Chapter X-A is Parliament’s answer to Vodafone on the first point: it supplies by statute the look-through that the Court held was a matter of policy. That is why arguments drawn from Vodafone about the limits of judicial anti-avoidance have less force inside Chapter X-A than outside it, and a submission that relies on Vodafone without acknowledging that should expect to be told so.

Does the judicial doctrine survive GAAR? The Telangana High Court in Ayodhya Rami Reddy Alla treated Chapter X-A as a structural codification of the earlier judicial rules and applied both, and the Supreme Court in Tiger Global is reported to have held that judicial anti-avoidance operates in parallel with the statute and can be used even where GAAR is inapplicable. That proposition has a real edge for taxpayers. If it is right, the Rule 10U gates are not a safe harbour at all: a tax benefit of two crore rupees, or income from an investment made in 2015, escapes Chapter X-A but not the doctrine. Whether a judicially developed rule can be applied to do what Parliament has set conditions for doing, and has declined to do below those conditions, is a question that has not been argued out, and it is the other large open question in this area. The Revenue’s answer to it is that the judicial doctrine is not a competing anti-avoidance power at all but a rule for construing the charging and computation provisions, so that the conditions in Rule 10U, which are conditions on the exercise of a statutory power, have nothing to say to it.

Dividend stripping before section 94(7): what Walfort decided

The closest analogue to bonus stripping decided by the Supreme Court is dividend stripping, and the decision is CIT v. Walfort Share and Stock Brokers P. Ltd., Civil Appeal No. 4927 of 2010, decided on 6 July 2010 by Kapadia CJI and Swatanter Kumar J, reported at (2010) 8 SCC 137 and (2010) 326 ITR 1.

The facts are the clearest statement of the technique in the decided cases. On 24 March 2000, the record date, the assessee bought units of a mutual fund at a net asset value of Rs 17.23 and received a tax-free dividend of Rs 4 per unit, Rs 1,82,12,862.80 in all. On 27 March 2000, the next trading day, the net asset value had fallen to Rs 13.23, exactly the dividend, and the assessee sold the whole holding. It claimed the dividend exempt and set off the resulting loss. The Assessing Officer disallowed the loss to the extent of the dividend as artificial and tax-motivated.

The Supreme Court held the disallowance wrong. The year was before section 94(7) applied, and the Court said:

“We are of the view that the AO had erred in disallowing the loss.”

“Hence, in the cases arising before 1.4.2002, losses pertaining to exempted income cannot be disallowed.”

“Even assuming that the transaction was pre-planned there is nothing to impeach the genuineness of the transaction.”

And on what Parliament had done about it:

“the Parliament has not treated the dividend stripping transaction as sham or bogus.”

“The object of Section 94(7) is to curb the short term losses.”

The case pulls in different directions, which is why it is worth reading rather than citing.

For the taxpayer. A deliberately planned loss, engineered to the rupee against a known corporate action, was held genuine and allowable before the specific rule applied to it, and Parliament’s response was a measured rule that ignores the loss only to the extent of the dividend rather than condemning the transaction.

For the Revenue. Walfort was decided in 2010, eight years before GAAR applied. It says nothing about Chapter X-A and it could not. Its holding that the loss is allowable is a holding about the computation provisions and section 14A, not about a chapter that empowers the Revenue to disregard a transaction notwithstanding anything in the Act. Section 95(1) was enacted precisely to change the position Walfort describes.

And for both. Walfort contains no sentence about bonus stripping. An order or a submission that attributes one to it is wrong, and the mistake is made often enough to be worth checking before it is relied on.

What the Board has said: Circular No. 7 of 2017

The Central Board of Direct Taxes issued sixteen clarifications on GAAR in question and answer form on 27 January 2017, shortly before the Chapter took effect. It will be produced across the table and a practitioner should know what is in it, with a caution about the source.

The caution first. The Department’s own published file of the circular is a scanned image without a text layer and could not be read for this article, and the departmental bulletin that reproduces it is not reachable. What follows is taken from the Press Information Bureau release of the same date, which is official but is a summary, and from independent reproductions of the full text which agree with each other. It should be checked against a printed copy before any of it is quoted in a submission.

The first answer is the one that matters most here, and it goes against the taxpayer. Question 1 asks whether GAAR will be invoked if a specific anti-avoidance rule applies. The Board’s answer is that it may:

“It is internationally accepted that specific anti avoidance provisions may not address all situations of abuse and there is need for general anti-abuse provisions in the domestic legislation. The provisions of GAAR and SAAR can coexist and are applicable, as may be necessary, in the facts and circumstances of the case.”

So the Board did not adopt the Shome Committee’s recommendation that a specific rule should displace the general one on the element it covers; it declined to. Anyone running the argument set out in Part 2 must meet that answer, and must meet it knowing that it is the Revenue’s own published position and not merely counsel’s. The answer is not fatal to the argument, because a circular cannot control the construction of a section and because the Board’s words still leave the question to “the facts and circumstances of the case”, which is precisely where the proportionality point about the consequence belongs. But an article that told you the Board was on your side here would be doing you no favours.

Three other answers bear directly on this subject.

Court and tribunal sanction, question 8. Asked whether GAAR will be invoked where an arrangement is sanctioned by a court, the National Company Law Tribunal or settled judicial precedent, the Board answered that “Where the Court has explicitly and adequately considered the tax implication while sanctioning an arrangement, GAAR will not apply to such arrangement.” Read the qualification: sanction alone is not enough, and the question in any scheme case is what the tribunal was actually shown about tax. That is directly in point on facts like EKGE’s, where the identity of the person who made the gain and the person who claimed the set-off is supplied by two amalgamation orders.

The threshold, question 14. The Board confirmed that the Rs 3 crore limit is tested by reference to the tax benefit arising from the arrangement, aggregated across the parties in a year, rather than taxpayer by taxpayer, which is how Rule 10U(1)(a) reads.

Consistency across years, question 15. Where the Approving Panel has held an arrangement permissible, the Board’s answer is that the Revenue will not ordinarily reopen that question for a later year on the same facts. That is worth having on a file where a technique recurs annually.

The circular also deals with grandfathering of bonus issues, splits and conversions, with leases and loans, with foreign portfolio investors and participatory notes, with the interaction between a treaty limitation of benefits article and the Chapter, with corresponding adjustments, and with a five-year immunity from penalty in certain cases. Those are outside the scope of this article and are listed so that the reader knows what else is in the document. There is nothing in it about the position of a deductor at the withholding stage.

The Department’s case, at its highest

It is worth setting out before the answers, because a taxpayer who has read only the decisions in his favour is not prepared. What follows is the case as a standing counsel would build it, not the case that was actually run in EKGE.

Start with the statute’s own instruction. Chapter X-A opens with a non obstante clause and section 100 says it applies in addition to or in lieu of any other basis for determining liability. Parliament enacted a general rule precisely because specific rules are always behind the planning, and the Board has said in terms that GAAR and a specific rule can coexist and are applied as the facts require. The taxpayer’s argument that a specific provision occupies the field is therefore contrary both to the text and to the Revenue’s published position, and has been rejected by the only court to decide it.

Then attack the construction, not the facts. The strict-construction canon on which the whole of EKGE rests comes from EKGE and from nowhere else. No decision of the Supreme Court applies such a canon to this Chapter, and a rule directed at avoidance which is read down against the Revenue defeats the purpose for which it was enacted. The definitions were drawn wide deliberately: “arrangement” extends to a step, to a part, and to an understanding that is not enforceable, and the Explanation to section 95 says the Chapter applies to a step as it applies to the whole.

And the threshold holding reads a requirement into the definition that is not there. Section 102(1) requires a transaction, operation, scheme, agreement or understanding, or a step in one. It nowhere requires that the outcome be determinate. The steps in a bonus stripping programme are entirely pre-ordained; what is not pre-ordained is how much the trade makes. If an arrangement needs a determinate outcome, then no unilateral multi-step restructuring carried out by one taxpayer at market prices can ever be an arrangement, which would take a large part of the Chapter out of use. Nor does the statute require two or more identifiable parties; that requirement is judicial gloss.

On purpose, the record supplies the admissions. The order itself records that the taxpayer was advised in advance that buying pre-bonus and selling post-bonus would yield a short-term return and that it invested with a view to achieving that objective. The purchases cluster before record dates, the sales follow the allotments, and the quantity sold matches the original holding. The written allocation was about Rs 50 crore to equities and Rs 106.31 crore went into pre-bonus shares. And the retention relied on so heavily is, on the face of section 94(8)(c), the parcel that the technique itself assumes a bonus stripper will hold.

Where the taxpayer’s own benefit is modest, section 99 answers it. Connected persons may be collapsed and an accommodating party disregarded for the very purpose of deciding whether a tax benefit exists, which is also how the arrangement is brought over the three crore line.

The limb that was never pleaded is the strongest one. Section 97(1)(d) asks whether the arrangement has a significant effect on the business risks or net cash flows of any party apart from the tax benefit. A fortnight’s market exposure on a matched purchase and sale is slight against a loss sheltering a gain many times its size. Section 97 was not argued in EKGE and clause (a) was not invoked, so the version of the Revenue’s case that won in Ayodhya Rami Reddy Alla has never been tested on exchange-traded facts.

There is now a Supreme Court decision, and the Revenue will open with it. Whatever the limits of Tiger Global as a determination under Chapter X-A, it is reported to hold that the onus shifts to the taxpayer and that judicial anti-avoidance doctrine survives alongside the statute. A High Court asked to construe the Chapter strictly against the Revenue will be shown it, and the answer that it was a decision on section 245R(2) has to be made rather than assumed. The Revenue’s reply to that answer is that a judgment of the Supreme Court expressed in impermissible-avoidance language binds under Article 141 whatever the jurisdictional frame in which it was given.

And the relief granted in EKGE is vulnerable on its own terms. A writ court decided purpose, which is a question of fact, on affidavit, and declared an entitlement to the set-off rather than remitting the matter to the body Parliament constituted to decide it. Where the complaint is that the Panel failed to engage with the record, the orthodox order is a remand so that it can. On a search file the Revenue adds that the reference was made inside a section 153A proceeding which had abated, so the incriminating-material argument is not available either.

That case failed on the EKGE facts, and it failed for reasons about evidence rather than sympathy: the Revenue identified no step whose main purpose it could demonstrate, left the commercial narrative unanswered, and advanced clauses (b) and (d) by pointing to the tax result rather than to the misuse or to the manner. But it failed as it was run. EKGE’s own best answer to the remand point, which a defender should have ready, is that where the Revenue has failed to discharge a burden on the material it chose to put forward there is nothing to remit, because it does not get a second opportunity to make out a case it did not make.

The clause the auditor has to answer first: Form 3CD clause 30C

Before any of the defence below is reached, a chartered accountant has usually already answered the GAAR question once, in a document signed and filed before any notice issued. This is the most immediate professional exposure in the whole subject and it is rarely discussed.

Clause 30C of Form No. 3CD, prescribed under rule 6G(2), reads:

“30C. (a) Whether the assessee has entered into an impermissible avoidance arrangement, as referred to in section 96, during the previous year? (Yes/No)

(b) If yes, please specify:—

(i) Nature of the impermissible avoidance arrangement:

(ii) Amount (in Rs.) of tax benefit in the previous year arising, in aggregate, to all the parties to the arrangement:”

It was inserted, together with clause 30A on secondary adjustments and clause 30B on the interest limitation, by the Income-tax (Eighth Amendment) Rules 2018 with effect from 20 August 2018.

And it is operative, whatever you may have been told. Reporting under clause 30C was kept in abeyance by four successive circulars: Circular No. 6/2018 of 17 August 2018 to 31 March 2019, Circular No. 9/2019 of 14 May 2019 to 31 March 2020, Circular No. 10/2020 of 24 April 2020 to 31 March 2021, and Circular No. 5/2021 of 25 March 2021 to 31 March 2022. There is no circular after that. Reporting has therefore been live for every tax audit report furnished on or after 1 April 2022. Commentary still describing the clause as deferred has been wrong for four years, and some of it cites Circular No. 2/2024 for a further deferral, which is a circular about Form 10B and Form 10BB for trusts and has nothing to do with Form 3CD. Of the circulars above, only the first could be read in an official source for this article, through the Press Information Bureau release announcing it; the other three are corroborated across professional sources including the Institute’s own Guidance Note, which names Circular No. 5/2021 as the operative deferral.

Read what the clause actually asks, because it is remarkable. It asks the auditor to say whether the assessee has entered into an impermissible avoidance arrangement “as referred to in section 96”. It does not mention section 144BA. It does not mention the Approving Panel. And it carries no threshold, where Rule 10U(1)(a) keeps the Chapter out altogether below three crore rupees, so the set of arrangements reportable under clause 30C and the set to which Chapter X-A applies are not the same set.

The difficulty follows from the section 144BA architecture described in Part 2. A declaration that an arrangement is impermissible can be made only through section 144BA, after a notice, objections and a hearing, by a three-member panel chaired by a serving or former High Court judge, and the resulting directions bind the Commissioner as well as the assessee. Neither the Assessing Officer nor the Commissioner can reach that conclusion alone. Clause 30C nevertheless asks a chartered accountant to answer it, in a form signed before any of that machinery has been engaged.

The Department’s own reference material does not engage the problem: it sets out four cumulative ingredients the auditor is to satisfy himself about, including that the arrangement is “impermissible”, and says nothing about section 144BA, the Approving Panel or Rule 10U. The Institute’s Guidance Note on tax audit does address clause 30C, and on the summaries available it directs the auditor to the narrower and more defensible enquiry of whether the Commissioner or the Approving Panel has already declared an arrangement impermissible or referred one for declaration, and to indicate in Form 3CA or 3CB where the aggregate tax benefit cannot be ascertained. The relevant pages of the Guidance Note could not be read for this article, so that account is taken from secondary summaries and the Guidance Note should be read before it is relied on. The Bombay Chartered Accountants’ Society has put the objection more bluntly, that a tax auditor “is not competent to say that a particular arrangement is an impermissible avoidance arrangement” because “it is only the Approving Panel which can declare an arrangement as impermissible avoidance arrangement”, and has proposed a disclaimer.

The practical course, pending anything authoritative, is a reasoned negative answer rather than a bare one: record in the working papers the basis on which the answer was given, including the Rule 10U computation showing the aggregate tax benefit, obtain a management representation, and consider a qualifying note where the position is genuinely open. An unqualified “No” that the Revenue later contradicts is the first document produced across the table. A “Yes” entered out of caution is an admission on the central issue, made by someone the statute does not authorise to make it.

One change worth knowing for tax year 2026-27. The successor to Form 3CD under the Income-tax Act, 2025 is Form No. 26, prescribed by rule 47 of the Income-tax Rules, 2026. Its Part D particulars carry forward the neighbours of clause 30C, the secondary adjustment becoming clause 40 and the interest limitation clause 41, but they contain no counterpart to clause 30C. On the notified form the GAAR reporting obligation in the tax audit report does not survive the move to the new Act. Note also that Form No. 26 has a clause numbered 44, which is unrelated to the old clause 44 on goods and services tax expenditure; that becomes clause 52(b).

Defending a GAAR reference

Roughly in this order. The first three are arithmetic and dates rather than argument, and they are capable of disposing of the file before any of the doctrine is reached. The provision numbers given are the 1961 Act’s, which govern the years now under assessment and appeal; for tax year 2026-27 onwards read section 179 for section 96, section 180 for section 97, section 274 for section 144BA and rule 128 for rule 10U, and the concordance table below for the rest.

Work Rule 10U before anything else. Compute the tax benefit on the Rule 10U(3)(iv) basis, which for an increase in loss is the tax that would have been chargeable had the increase been the total income, and aggregate it across all the parties for the relevant assessment year, taking “party” from section 102(6) and bearing in mind that section 99 lets connected persons be collapsed. If the aggregate is three crore rupees or less, Chapter X-A does not apply. Then check when the investment was made: if the income is from the transfer of an investment made before 1 April 2017, clause (d) is engaged, and check which version of clause (d) and sub-rule (2) governs the year.

Check the year. Chapter X-A applies only from assessment year 2018-19. For an earlier year the Revenue is on the judicial doctrine, and an order that uses section 95 or section 97 as an aid to construction in a pre-GAAR year should be challenged for doing so rather than answered on its merits.

On a search file, settle abatement first. Where the reference is made inside a section 153A proceeding, whether that assessment can carry any addition at all depends on whether it had abated, and that question comes before everything in these three Parts.

Answer the section 144BA(2) notice, whatever the state of the objections. Silence costs the Approving Panel altogether, because the reference under sub-section (4) arises only where the assessee has objected. This is the single most consequential procedural step in the chapter and it is lost by inaction.

Force the Revenue to name the clause. Rule 10UB(2)(iv) requires the pre-reference notice to state “the basis and the reasons why the arrangement satisfies the condition provided in clause (a), (b), (c) or (d) of sub-section (1) of section 96”, and sub-clause (v) requires the documents and evidence relied on for it. A notice that asserts an impermissible avoidance arrangement without identifying the clause has not complied, and the omission matters: the four clauses require different evidence, and a taxpayer who does not know which is alleged cannot answer it. Check also whether the notice was issued under Rule 10UB(1) at all, or whether it was a section 142(1) questionnaire later described as one.

Separate the two conditions in your reply, and answer them separately. Deal with the threshold under section 102(1) first, then the main purpose, then the identified clause. Do not run them together, because the Revenue’s characteristic error is to run them together and the defence depends on keeping them apart. Open on the strict construction holding in EKGE, which is the answer to the width of the definitions.

Answer the main purpose question with contemporaneous documents, not with assertions. What won EKGE was a record: advice taken from named institutions, an internal consultant retained, funds parked in liquid instruments pending a decision and returning income that was itself offered to tax, a written allocation between asset classes, and a rationale for the trade that existed before the trade. Board minutes, investment committee notes, advisers’ letters, the mandate given to a portfolio manager and the research note on which the trade was placed are the material. A narrative constructed after a section 144BA(2) notice is worth very little and looks like what it is. Where the execution departed from the written plan, explain the departure in the same document set rather than leaving the Revenue to find it.

Put the group’s own assessment history on the record. Where the reference follows a search or a group-wide exercise, tabulate the other assessments and their outcomes. The comparison was an independent ground in EKGE and it is discoverable material the taxpayer already has.

Meet section 97 even if it has not been pleaded. It is the limb on which the one adverse decision was given. Be ready on duration with section 97(4)(i), which provides that the period for which an arrangement exists “may be relevant but shall not be sufficient”; on offsetting elements under section 97(1)(b)(iii); and on business risks and net cash flows under section 97(1)(d) with the actual price movement over the holding period, including the trades that lost money.

Put every point to the Commissioner, not to the Panel. He is the only person in the chain who can stop the matter under section 144BA(5), and he must act within the Rule 10UC limits. Treat the objections as the main submission of the case, annex the paper book, and ask for the personal hearing that sub-section (2) expressly provides for, recording the request in the objections themselves.

Prepare the Panel hearing as the only oral hearing you will get. Ask for the hearing expressly under section 144BA(7), recording that any adverse direction would be prejudicial. File a paper book. If an inquiry is directed under sub-section (8), ask for the report and for an opportunity to answer it. And remember sub-section (11): if other years are in issue, address them now, because directions can be made to govern years for which no fresh reference is needed.

Get the dates. Ask in writing for Form 3CEG, Form 3CEI and the date on which the Commissioner’s reference was received by the Panel. Without that last date the six-month limit in section 144BA(13) cannot be computed, and whether a direction issued outside it is void is an open question worth preserving.

Deal with recovery, and know who can appear. A GAAR reference does not itself create a demand; the demand arises on the assessment made in conformity with the directions, and from that point recovery runs as it does on any other demand, so a stay application belongs with the appeal to the Tribunal under section 253(1)(e) and the usual deposit practice applies. The rest of the assessment is not stayed while the reference is pending; section 144BA(1) permits the reference at any stage and section 153 simply excludes the detour from the limitation, so the officer continues with the other issues. On representation, a chartered accountant is an authorised representative for proceedings under the Act and the Panel’s own procedural rules could not be read for this article, so confirm the position before the hearing rather than at it.

Keep the remedy question in view from the start. If the directions go against you, the choice is a writ under Article 226 or an appeal against the resulting assessment direct to the Tribunal under section 253(1)(e). A writ is the stronger course where the complaint is that the Panel applied the wrong test or never addressed an ingredient, and the weaker course where the complaint is about the evaluation of evidence. Frame the objections and the hearing so that the record supports the former, and if you go to the writ court ask for a declaration of entitlement rather than a remand, which is what EKGE gave.

And tell the client what the exposure actually is. Not simply the denial of the claim: recharacterisation under section 98, which can produce a liability the transaction never carried; no corresponding adjustment, so the same amount can be taxed twice across a group; the treaty displaced by section 90(2A); other years reached under section 144BA(11); and a penalty case under section 270A with no section 270AA immunity if misreporting is alleged. Invoke Rule 10UA expressly to confine the consequences where only part of an arrangement is impugned.

Structuring so that the question does not arise

For a transaction being planned now the governing text is the Income-tax Act, 2025, so the numbering below is the new one with the familiar one in brackets.

Start by asking whether a specific rule already covers the ground. If it does, the consequence Parliament prescribed is the measure of the mischief and the Revenue has to justify anything harsher; if it does not, and the conduct is of a kind Parliament has since brought within a specific rule, the Shome Committee argument is available, though the Board’s own clarification of 27 January 2017 is against it. The specific anti-avoidance provisions to run the transaction against are listed in Part 2, in their 1961 numbering; their counterparts in the 2025 Act were not traced for this article.

Then run the three crore test properly, under rule 128(1)(a) of the Income-tax Rules, 2026 (rule 10U(1)(a) of the 1962 Rules). It aggregates across all the parties, not the client, and section 182 (section 99) can bring connected persons inside that aggregate. Model the group’s benefit and not the entity’s, and model it year by year, because the test is annual. Where the benefit is genuinely below the line the Chapter does not apply at all, and that is a threshold Parliament enacted rather than something to be embarrassed about. The caution is that spreading an arrangement across years is a different thing from a single year falling below the line, and the Revenue will look at the arrangement as a whole.

Expect no advance clearance, because there is none. Chapter XI, like Chapter X-A before it, contains no ruling procedure, no safe harbour and no analogue of an advance pricing agreement. The only route to advance certainty is an application to the Board for Advance Rulings, and the Tiger Global litigation shows the difficulty with it: a prima facie view that the transaction is designed for the avoidance of tax is itself a ground on which the application can be rejected, under what was proviso (iii) to section 245R(2), so the very feature that makes an applicant want certainty may be the feature that denies it. The practical consequence is that the file built before the transaction is the only protection available, and it is built for a reader who will be hostile.

Give each step its own reason, not just the transaction. This is the specific risk created by section 178(2) (the Explanation to section 95) and the presumption in section 179(2) (section 96(2)). A commercially impeccable transaction containing one tax-driven step is not protected by its own commerciality; the step is the unit of analysis and the presumption then colours the whole. Any step whose only explanation is the tax result should be removed or given one.

Document the alternatives, not just the decision. A board minute that records the course chosen is evidence of a decision. A board minute that records the options considered, including a less tax-efficient one, and the commercial reasons for preferring the one adopted, is evidence of a purpose, which is what section 96(1) actually asks about. The document should be contemporaneous with the decision and not with the transaction.

Do not make the tax the only return. The bonus arbitrage in EKGE had an expected commercial return the Court recited and could test. A trade whose only expected return is the tax saved has no answer to section 179(1), formerly section 96(1), and none to section 180(1)(d), formerly section 97(1)(d), either.

Watch the terms, not just the price, in anything intra-group. Clause (a) of section 179(1), formerly section 96(1)(a), is the limb a related-party transaction will be tested against, and an arm’s length price certified under the transfer pricing code is not by itself an answer to a right or obligation that no independent party would have accepted. Tenure, security, conversion and exit terms, voting and ratchet arrangements, waivers and subordinations, and swap ratios should each be capable of being defended as terms an outsider would have agreed.

Transact at market, through ordinary channels, and pay the transaction taxes. That is not a safe harbour, and no statute says it is, but it is the factual difference between the decision that went against the taxpayer and the two that went for him. The corollary is the real advice: a leg routed to a related party, or a price the parties set between themselves, converts a defensible position into a difficult one, and it is usually inserted for convenience rather than necessity.

Avoid circularity of funds. Round trip financing is a deeming provision under section 180(2), formerly section 97(2), the definition is inclusive rather than exhaustive, and it is determined without regard to whether the funds can be traced, to the sequence of transfers or to the means used. Funding an acquisition from a group entity that is itself funded by the proceeds is the fact pattern that lost Ayodhya Rami Reddy Alla.

Keep a position where the technique allows one, and be able to say why that quantum. Retention of the parcel the technique requires you to hold proves nothing. Retention beyond it, at a level that can be explained on investment grounds, is the fact that answers the purpose question.

And where a treaty is in the picture, do not treat it as the answer. Section 90(2A) disapplies the beneficial-provision rule so far as Chapter X-A is concerned; under the new Act the treaty power is section 159 and the override is at section 159(6).

Nor take comfort from “main purpose” where a treaty is involved. The domestic rule asks whether the main purpose of the arrangement was a tax benefit. The principal purpose test imported into most of India’s treaties through the multilateral instrument asks whether obtaining the benefit was one of the principal purposes, which is a materially lower threshold, and it is applied by the treaty partner’s authorities as well as by ours. A structure that survives section 179(1) can still fail the treaty test, and the two enquiries are not conducted by the same people or at the same time. A limitation of benefits article is worth having, both for its own sake and because it supports the argument that the contracting states measured the mischief themselves, but it is an argument and not a shield.

The Income-tax Act, 2025: the full concordance, and what actually changed

The Income-tax Act, 2025 came into force on 1 April 2026 and governs tax year 2026-27 onwards, so for anything being planned now it is the operative statute and the 1961 Act is the one that governs the years still under assessment and appeal. Both are therefore live, which is why these three Parts give the 2025 number alongside each provision as they go. For anyone holding both in mind, the chapter was re-enacted rather than rewritten: the main purpose test, the four limbs, the presumption, the deeming provisions on commercial substance and the consequences all survive in substantially identical words.

One point of vocabulary first, because the labels collide. Assessment year 2026-27, that is the financial year 2025-26, is the last year governed by the 1961 Act and is the year most files now in hand concern. Tax year 2026-27, the year beginning 1 April 2026, is the first governed by the 2025 Act. The two carry the same numbers and are a year apart, so a note on a file should say which Act it is using. “Notwithstanding” becomes “irrespective of” throughout, which is the new Act’s drafting convention and not a change of meaning, and “previous year” becomes “tax year”.

The concordance

Income-tax Act, 1961	Subject	Income-tax Act, 2025
Chapter X-A, ss. 95 to 102	General anti-avoidance rule	Chapter XI, ss. 178 to 184
s. 95(1)	Applicability, non obstante	s. 178(1)
s. 95(2)	Applies from assessment year 2018-19	no counterpart, spent
s. 95, Explanation	Application to a step or part	s. 178(2)
s. 96(1)(a) to (d)	Impermissible avoidance arrangement	s. 179(1)(a) to (d)
s. 96(2)	Presumption from a tainted step	s. 179(2)
s. 97(1), (2)	Deemed lack of commercial substance; round trip financing	s. 180(1), (2)
s. 97(3)	Accommodating party	s. 184(1), not s. 180(3)
s. 97(4)	The “may be relevant but shall not be sufficient” list	s. 180(3)
s. 98(1)	Consequences, the seven listed	s. 181(1) and (2)
s. 98(2)	Equity as debt, capital as revenue	s. 181(3)
s. 99	Connected persons and accommodating parties	s. 182
s. 100	“In addition to, or in lieu of”	s. 183(a)
s. 101	Guidelines and conditions as prescribed	s. 183(b)
s. 102	Definitions	s. 184, every clause shifted by one from “arrangement”
s. 102(1) arrangement		s. 184(2)
s. 102(2) asset		s. 184(3)
s. 102(3) benefit		s. 184(4)
s. 102(4) connected person		s. 184(5)
s. 102(5) fund		s. 184(6)
s. 102(6) party		s. 184(7)
s. 102(7) relative		s. 184(8), pointing to s. 92(5)(g)
s. 102(8) substantial interest		s. 184(9)
s. 102(9) step		s. 184(10)
s. 102(10) tax benefit		s. 184(11)
s. 102(11) tax treaty		s. 184(12), pointing to s. 159
s. 144BA	Reference, Commissioner and Approving Panel	s. 274, with 24 sub-sections
s. 144BA(12)	Prior approval before the order	s. 274(12)
s. 144BA(13)	Six months for the Panel	s. 274(13)
s. 144BA, Explanation	Exclusions and the sixty-day floor	s. 274(14) and (15)
s. 144BA(14)	Binding effect; bar on appeal	s. 274(16) and s. 274(17)
s. 144BA(15), (16)	Constitution and composition of the Panel	s. 274(18), (19)
s. 144BA(19)	Borrowed advance-ruling powers	s. 274(22)
s. 144BA(21)	Rule-making power	s. 274(24)
s. 253(1)(e)	Direct appeal to the Tribunal	s. 362(1)(e)
s. 245U	Powers borrowed by the Panel	s. 387, Board for Advance Rulings
s. 90, s. 90A	Treaties, and the s. 90(2A) override	s. 159
Rule 10U	Chapter not to apply in certain cases	Rule 128 of the Income-tax Rules, 2026
Rule 10UA	Consequences of a part arrangement	Rule 127
Rule 10UB	Notice and forms for the reference	Rule 129
Rule 10UC	Time limits	Rule 130
Forms 3CEG, 3CEH, 3CEI	Reference, dropping, satisfaction	Forms 62, 63, 64

Two cautions about reading that table. The rules invert: rule 127 of the 2026 Rules is the consequences rule and rule 128 is the non-application rule, the reverse of 10U and 10UA, so do not assume 10U becomes 127. And the sub-section mapping within section 274 is set out above only where it has been checked against the text on both sides; for sub-sections (1) to (11) the content matches in the same order but the alignment is inference from matching content, so cite section 274 generally rather than a sub-section unless you have read it.

No provision outside Chapter X-A and section 144BA has been traced into the 2025 Act for this article, apart from section 253(1)(e), section 245U and the treaty provisions, which appear in the table above. So where a 1961 provision is discussed in these three Parts without a 2025 number beside it, including sections 70, 74, 55, 45(2A), 50B, 94, 246A, 270A, 153A, 144C, 245R and the specific anti-avoidance provisions listed earlier, that is why: the counterpart was not checked and is not guessed at here.

What actually changed

Change	Where	What it does
The accommodating party definition moves	from s. 97(3) to s. 184(1)	s. 180 has three sub-sections where s. 97 had four, so s. 180(3) is the counterpart of s. 97(4), not of s. 97(3); citing s. 180(3) for the accommodating party definition would be wrong
The assessment year switch-on disappears	s. 95(2) has no counterpart	spent: the new Act’s own commencement does that work. The gates are unaffected, because they never lived in s. 95; the Rs 3 crore threshold and the 1 April 2017 grandfathering are in rule 128
The appeal bar separates from the binding provision	s. 144BA(14) becomes s. 274(16) and s. 274(17)	citing “section 274(16)” for the appeal bar is wrong
The Explanation becomes operative sub-sections	the Explanation to s. 144BA becomes s. 274(14) and (15)	the exclusions and the sixty-day floor move into the body of the section
The borrowed powers change their source	s. 144BA(19) becomes s. 274(22)	the Panel takes the powers of the Board for Advance Rulings under s. 387, not of the Authority for Advance Rulings under s. 245U
The removal-of-doubts preamble goes	s. 97(4) becomes s. 180(3)	the operative words “may be relevant but shall not be sufficient” are unchanged, but the clause is no longer expressed as clarificatory
The recharacterisation lead-in widens	s. 98(2) becomes s. 181(3)	“For the purposes of sub-section (1)” becomes “In this section”, which on its face extends the equity-for-debt and capital-for-revenue powers to the whole section rather than to its opening words

One question the concordance does not answer and a planner has to. Section 95(2) fixed the switch-on at assessment year 2018-19 and has no counterpart, because the new Act’s own commencement does that work. What Chapter XI does not address anywhere is the arrangement entered into under the old Act whose tax benefit falls in a year governed by the new one. On the rules the position looks continuous: rule 128 carries both gates forward and the grandfathering is keyed to when the investment was made rather than to which Act was in force. But the repeal and savings provision of the 2025 Act was not traced for this article, and anyone structuring across the changeover should read it rather than assume continuity from the rules.

Two of the changes above are worth a sentence more.

The recharacterisation change is probably of no practical effect, because the list of consequences is itself the elaboration of the opening words, but it is a textual broadening rather than pure renumbering and a careful reader will notice it.

And the new Act did not take the opportunity to soften “the main purpose” to “one of the main purposes”, which is the formulation used in the principal purpose test in the multilateral instrument and in several foreign general anti-avoidance rules. India’s general rule remains a single-main-purpose test, which, together with the three crore exclusion in the rules, is the most significant protection the chapter contains.

A note for anyone working from these orders

Five things could not be closed against a primary source. Each is flagged where it arises and they are gathered here so they can be found: the primary text of Anvida Bandi; the operative half of the Supreme Court’s judgment in the Tiger Global matter and its concurring opinion; the official text of CBDT Circular No. 7 of 2017; the final version of the Shome Committee report, and with it Example 9; and the rules governing the Approving Panel’s own procedure. Three smaller ones belong with them: the contents of Form No. 3CEG, the pages of the Institute’s Guidance Note dealing with clause 30C, and three of the four circulars deferring that clause.

One of those deserves more than a cross-reference, because the whole comparison between the Telangana decisions turns on it. It could not be established whether Anvida Bandi involved a bonus issue at all. It is said to have concerned shares of a large listed technology company which did make a one for one bonus issue with a record date falling in the relevant year, and the figures are consistent with a bonus pattern; but they are equally consistent with the market fall of early 2020, and no available account states it. If that case was a bonus stripping case, the silence about Ayodhya Rami Reddy Alla is a serious criticism of it. If it was not, the two are less comparable than they look, and the reconciliation of the three decisions in Part 2 rests on two of them rather than three.

And some defects on the face of the orders should be known before either is quoted.

In Ayodhya Rami Reddy Alla the bonus ratio is given as 5:1 at paragraph 4 and as 1:5 at paragraph 22, and section 144BA is rendered as “144AB” three times, twice at paragraph 38 and once at paragraph 43.

In EKGE Retail there are five. Paragraph 29 describes Chapter X-A as “introduced w.e.f. 01.04.2018”, which is accurate as to when the Chapter applies under section 95(2) but not as to when it was inserted, which was 1 April 2016. Paragraph 18 describes the Assessing Officer’s reference to the Commissioner as having been made “under Section 144BA(2)”, where sub-section (1) is the reference provision and sub-section (2) is the Commissioner’s notice. Paragraph 20 twice refers to “Section 70(2)(i)”, though sub-section (2) as substituted by the Finance Act 2002 has no clauses; the Court reverts to “Section 70(2)” later. The Revenue’s submission at paragraph 26 is recorded as resting on “Section 55(a)(iiia)”, where the provision is section 55(2)(aa)(iiia). And the cause title is “EKGE Retail LLP” while paragraph 5 records the merger as having been into “EKGE Retail Private Limited”, with paragraph 6 describing Madevsag Retail as “the petitioner herein”; anyone drafting a cause title or a citation should notice that before copying it.

Two further slips are worth flagging because a defender might otherwise quote them. The worked bonus illustration at paragraph 10 does not produce the return it states: on the figures given, a holding worth Rs 1,000 before a one for one bonus becomes two shares at Rs 501, which is Rs 1,002 and a return of 0.2 per cent, not the 2 per cent the paragraph asserts and not the 50 per cent annualised that follows from it. The commercial rationale survives the arithmetic, because an arbitrage of that kind is real and is measured in fractions of a per cent, but the figures as printed do not work. And the sentence at paragraph 21 recording the petitioner’s submission, “The price of the stock in questions was based entirely on market improbable”, is garbled and should not be quoted.

Where this leaves the subject

The striking thing about Indian GAAR so far is how little of the litigation is about avoidance. The decided cases turn on the structure of a definition, on who has to prove what and in what order, on whether a threshold in a rule was crossed, and on whether a High Court can decide a disputed fact in a writ. Those are not the questions the chapter was enacted to raise, and the reason they dominate is that Chapter X-A is unusually demanding of the officer who invokes it. It requires an arrangement, a main purpose, an identified clause, a tax benefit above a stated figure, and a hearing before a judicially chaired panel, and an order that skips any of those is vulnerable however unattractive the planning looks.

For the Revenue the lesson of the Telangana line is that the chapter rewards evidence and punishes characterisation. An order that reasons from the tax result to the taxpayer’s purpose will not survive, and an order that does not say which of the four clauses is relied on has not begun. For the taxpayer the lesson is the mirror image: the material that answers a GAAR allegation is material that exists before the allegation, and a reason recorded at the time, with the alternatives that were rejected, is worth more than any submission made afterwards.

Two things will decide how this field develops and neither is in anyone’s hands at the moment. The appeal in Ayodhya Rami Reddy Alla is pending before the Supreme Court with leave granted and an interim order in the taxpayer’s favour, and it squarely raises whether a general anti-avoidance rule can be applied to conduct that Parliament chose to address, later and specifically, elsewhere. And the operative reasoning of the Supreme Court’s own GAAR judgment of January 2026 is not freely readable, so what it decides about Rule 10U and about the burden of proof is at present known mainly at second hand.

In the meantime the practical position is narrow and worth stating plainly. Exchange-traded loss harvesting, supported by a contemporaneous commercial rationale and a retained position, has twice been held outside Chapter X-A by a High Court, on reasoning that has not been tested against section 97 or clause (a) and has not been tested on appeal. Loss harvesting routed through related parties with circulating funds has once been held inside it. From assessment year 2023-24 the question largely falls away for bonus stripping, because section 94(8) now reaches securities and the specific rule applies of its own force. Not entirely, though: section 100 still permits the Chapter to be applied in addition to the specific rule, and the Revenue may want the harsher consequence that section 98 allows and section 94(8) does not. What remains is everything else the chapter can be pointed at, and there the first question to ask of any order is the one the Telangana High Court asked and the Revenue could not answer: which clause, and on what material.

The other two Parts. Part 1: what makes an arrangement impermissible under section 96 sets out the Rule 10U gates, the two conditions in section 96(1), the four clauses and the section 102 definitions. Part 2: consequences, the section 144BA machinery and the remedy covers section 99 and the three crore aggregate, section 98, the section 100 question, the Approving Panel, the appeal position and the three Telangana High Court decisions.

This note is general commentary on the law as at 18 October 2026 and is not advice on any matter. The position in a particular case depends on its own facts.