Kompetenz-kompetenz is the rule that an arbitral tribunal may rule on its own jurisdiction, and separability of the arbitration clause is the rule that the clause is treated as an agreement independent of the contract containing it. Both sit in Section 16(1) of the Arbitration and Conciliation Act, 1996, and together they settle who rules first when a party challenges the tribunal’s authority, and how far a defect in the main contract reaches the clause. Indian law has pushed both further than the UNCITRAL Model Law it borrowed them from, most recently in SBI General Insurance v. Krish Spinning (2024), which discarded the “eye of the needle” test as still too intrusive. Both doctrines stop at the same place, which is an attack aimed at the formation of the arbitration agreement itself.
This article sets out kompetenz-kompetenz and separability of the arbitration clause, how the two doctrines support each other, and where each one stops.
The two are usually named in the same breath, and the habit hides a distinction worth keeping. Kompetenz-kompetenz allocates authority: it decides who rules on a jurisdictional objection, and in what order. Separability is a rule about the clause itself: the clause is a distinct agreement, so the fate of the contract around it is not automatically its own fate.
Section 16(1) carries both, but not equally. Kompetenz-kompetenz is the main limb. Separability arrives in clauses (a) and (b), introduced by the words “for that purpose”, and those three words have produced more argument than the rest of the provision.
The law has moved since most accounts of it were written. A seven-judge bench restated both doctrines in December 2023, a three-judge bench in 2024 discarded a referral test the Supreme Court itself had created fifteen months earlier, and judgments from January and February 2026 mark the current outer edges.
Kompetenz-kompetenz in Indian arbitration
Kompetenz-kompetenz in Indian arbitration is the rule that an arbitral tribunal may rule on its own jurisdiction, including on objections going to the existence or validity of the arbitration agreement that created it. The rule sits in the opening words of Section 16 of the Arbitration and Conciliation Act, 1996. A tribunal facing the objection “you have no authority to hear this” does not stop and wait for a court to answer. It answers first.
The provision states the competence in deliberately wide terms. The tribunal may rule “on its own jurisdiction, including ruling on any objections with respect to the existence or validity of the arbitration agreement”. Note what that sweeps in: not merely whether a dispute falls inside an admitted clause, but whether the clause exists at all. The mechanics of the remaining subsections, the deadlines for raising a plea and the route for challenging a ruling, are set out in our guide to Section 16 of the Arbitration and Conciliation Act, 1996, and the earlier treatment of the doctrine of kompetenz-kompetenz covers the ground before the 2023 and 2024 rulings reshaped it.
A rule of chronological priority, not of finality
The doctrine India adopted is a narrower thing than the doctrine its German name describes. Explaining the origin in In Re: Interplay Between Arbitration Agreements under the Arbitration and Conciliation Act 1996 and the Indian Stamp Act 1899, 2023 INSC 1066, a seven-judge bench, Chandrachud CJI recorded at paragraph 115 that the doctrine “as originally developed in Germany, was traditionally understood to imply that arbitrators are empowered to make a final ruling on their own jurisdiction, with no subsequent judicial review of the decision by any court”. India did not take that version. Here, the tribunal rules first and a court reviews later.
The cleanest statement of what that leaves is now more than twenty years old. In S.B.P. & Co. v. Patel Engineering Ltd., (2005) 8 SCC 618, Thakker J wrote at paragraph 22 that Section 16(1) “recognizes and enshrines an important principle that initially and primarily, it is for the Arbitral Tribunal itself to determine whether it has jurisdiction in the matter, subject of course, to ultimate court-control. It is thus a rule of chronological priority.”
That phrase carries two cautions. It comes from Thakker J’s separate opinion rather than the majority, and it is regularly quoted without that qualification. And SBP’s own holding on the referral court’s powers did not survive: the 2015 amendment to Section 11 was enacted to displace it, as the discussion below sets out. What survived is the definition. Kompetenz-kompetenz sequences the decision-makers. It does not remove one of them.
What the tribunal may rule on
The competence covers every jurisdictional question, not a shortlist. In Adavya Projects Pvt. Ltd. v. M/s Vishal Structurals Pvt. Ltd., 2025 INSC 507, decided on 17 April 2025, Narasimha J put the scope at paragraph 24:
“Section 16 embodies the doctrine of kompetenz-kompetenz, i.e., that the arbitral tribunal can determine its own jurisdiction. The provision is inclusive and covers all jurisdictional questions, including the existence and validity of the arbitration agreement, who is a party to the arbitration agreement, and the scope of disputes referrable to arbitration under the agreement.”
Adavya Projects is unusual and instructive, because the party the Supreme Court corrected was the tribunal. The tribunal had declined to decide whether certain parties were bound, treating the question as one that did not arise before it at that stage. At paragraph 28 the Court held that this “clearly shows that it did not exercise its jurisdiction in accordance with the principle of kompetenz-kompetenz”, and that the approach was “incorrect”. The competence is not a shield a tribunal may decline to pick up.
By 2025 the doctrine had stopped being contentious as a matter of principle. In Arabian Exports Private Ltd. v. National Insurance Co. Ltd., 2025 INSC 630, decided on 6 May 2025, Bhuyan J recorded at paragraph 40 that “the doctrine of Kompetenz-Kompetenz is now firmly embedded in the arbitration jurisprudence in India”, resting on the tribunal’s competence to rule on its own jurisdiction “including on the issue of existence or validity of an arbitration agreement”, with the object of minimising judicial intervention as “an acknowledgment of the concept of party autonomy”.
Separability of the arbitration clause
The separability of the arbitration clause is the rule that the clause is treated as an agreement independent of the contract it sits inside, so that a defect in the contract does not automatically become a defect in the clause. Section 16(1) states it in two parts. Clause (a) provides that “an arbitration clause which forms part of a contract shall be treated as an agreement independent of the other terms of the contract”. Clause (b) adds that “a decision by the arbitral tribunal that the contract is null and void shall not entail ipso jure the invalidity of the arbitration clause”.
Read those together and the practical effect is plain. One document contains two agreements: a bargain about goods, shares or services, and a separate bargain about how to resolve disputes over the first. The second does not die because the first does.
An agreement independent of the contract that contains it
The best Indian statement of why the rule exists came in Enercon (India) Ltd. v. Enercon GmbH, (2014) 5 SCC 1. Writing for the Court, Nijjar J put it at paragraph 83 of the report:
“The concept of separability of the arbitration clause/agreement from the underlying contract is a necessity to ensure that the intention of the parties to resolve the disputes by arbitration does not evaporate into thin air with every challenge to the legality, validity, finality or breach of the underlying contract.”
The same judgment records that Section 16 “accepts the concept that the main contract and the arbitration agreement form two independent contracts”. Without the rule, any party wanting to escape arbitration would have a standing move available: attack the contract, and the clause falls with it.
The current statement is SBI General Insurance Co. Ltd. v. Krish Spinning, 2024 INSC 532, decided on 18 July 2024. Pardiwala J framed the doctrine at paragraph 49 as a presumption by which “the arbitration agreement, by virtue of the presumption of separability, survives the principal contract in which it was contained”, and grounded it at paragraph 51 in what the parties were doing when they signed. The clause is included “with the mutual intention to settle any disputes that may arise under or in respect of or with regard to the underlying substantive contract, and thus by its inherent nature is independent of the substantive contract”.
Where the doctrine came from
Separability reached Indian law through the UNCITRAL Model Law, but its roots are older, and three points of ancestry are routinely reported wrongly.
The usual starting point is Heyman v. Darwins Ltd, [1942] AC 356, where the House of Lords held that repudiation does not abrogate the contract. Lord Macmillan put it at page 374: the contract “survives for the purpose of measuring the claims arising out of the breach, and the arbitration clause survives for determining the mode of their settlement. The purposes of the contract have failed, but the arbitration clause is not one of the purposes of the contract.” Heyman is the doctrine’s ancestor rather than its statement, and the seven-judge bench described it accurately at paragraph 95 of In Re Interplay as the decision the presumption “developed gradually in English law starting with”. The word “separable” does not appear in Heyman at all.
The American contribution is Prima Paint Corp. v. Flood & Conklin Mfg. Co., 388 U.S. 395 (1967), where Fortas J held for the Court that a claim of “fraud in the inducement of the arbitration clause itself” may be decided by a court, while claims of “fraud in the inducement of the contract generally” may not. The well-known description of arbitration clauses as “separable” from the contracts in which they are embedded appears in that judgment as a quotation of the Second Circuit’s reasoning, which the Court then adopted, rather than as the Supreme Court’s own coinage.
The modern English statement is Premium Nafta Products Ltd. v. Fili Shipping Co. Ltd., [2007] UKHL 40, and the name matters. Fiona Trust & Holding Corp v. Privalov is the Court of Appeal decision below, reported at [2007] EWCA Civ 20; the House of Lords decision carries the Premium Nafta title, though both names circulate for it. Lord Hoffmann stated the principle at paragraph 17: separability “means that the invalidity or rescission of the main contract does not necessarily entail the invalidity or rescission of the arbitration agreement”, and the arbitration agreement “can be void or voidable only on grounds which relate directly to the arbitration agreement”.
What the clause survives
Separability does its work when the main contract is attacked, terminated or discharged. The clause survives repudiation and breach, on Heyman’s reasoning. It survives a declaration that the contract is null and void, because Section 16(1)(b) says so in terms. It survives a defective main contract: the position on unstamped agreements, worked through in our analysis of the unstamped agreement question in Indo Unique Flame, is now that non-stamping is a curable defect rather than a ground of invalidity.
Krish Spinning added the commercially significant case, which is settlement. Parties routinely sign a discharge voucher or a full and final settlement and then argue about whether it was extracted under duress. Pardiwala J held at paragraph 53 that “even if the contracting parties, in pursuance of a settlement, agree to discharge each other of any obligations arising under the contract, this does not ipso facto mean that the arbitration agreement too would come to an end, unless the parties expressly agree to do the same”. At paragraph 115 the Court put the conclusion directly: the clause “continues to remain in existence even after the original contract stands discharged by ‘accord and satisfaction’”.
That reasoning was applied within a year. Arabian Exports concerned a flood loss claimed at about Rs 5.71 crore, settled by the insurer at about Rs 1.88 crore against a full and final discharge voucher, with the insured alleging economic duress. Bhuyan J held at paragraph 41 that whether the voucher was signed under duress, and whether the claim survived it, “are clearly within the domain of the arbitral tribunal”.
One qualification runs underneath all of this. Separability presupposes that a clause was validly formed in the first place. It protects an agreement that exists from the misfortunes of its host; it does not conjure one into being. Whether the parties made a valid arbitration agreement under Section 7 is a prior question, and the section below on the limits of the doctrine turns on it.
How the two doctrines hold each other up
The two doctrines hold each other up because each is unstable alone: separability gives the tribunal something to exercise jurisdiction under when the contract is attacked, and kompetenz-kompetenz gives separability an adjudicator willing to apply it. The seven-judge bench in In Re Interplay set out the link at paragraph 116, in the passage that has since become the standard citation:
“The ability of an arbitral tribunal to determine its own jurisdiction is an important facet of arbitration jurisprudence because it gives effect to the separability presumption. The separability presumption insulates the arbitration agreement from the defects of the underlying contract, and thereby ensures the sustenance of the tribunal’s jurisdiction over the substantive rights and obligations of the parties under the underlying contract even after such a contract is put to an end.”
Why a tribunal is not sawing off the branch it sits on
Without separability, a tribunal asked to declare the contract void would face a logical trap. Its authority comes from a clause inside that contract, so a ruling that the contract is void would appear to destroy the source of its own power to rule. The award would undo its own foundation.
Separability closes the trap by locating the tribunal’s authority somewhere the ruling does not reach. In Vidya Drolia v. Durga Trading Corporation, (2021) 2 SCC 1, Khanna J made the point at paragraph 115 of the report, in language that names the problem. The principle “prevents boot-strapping as it is primarily for the arbitral tribunal and not for the court to decide issues of existence, validity and rescission of the underlying contract”, and separation “authorises an arbitral tribunal to rule and decide on the existence, validity or rescission of the underlying contract without an earlier adjudication of the questions by the referral court”.
The consequence had been identified in A. Ayyasamy v. A. Paramasivam, (2016) 10 SCC 386. Chandrachud J recorded, at paragraph 4 of his concurring opinion, that severability “is a doctrinal development of crucial significance. For, it leaves the adjudicatory power of the arbitral tribunal unaffected, over any objection that the main contract between the parties is affected by fraud or undue influence.”
“For that purpose” and what it does not mean
Section 16(1) introduces separability with the words “for that purpose”, and those words invite a reading the Supreme Court has considered and declined. The natural inference is that separability exists only to serve kompetenz-kompetenz, as a device for allocating jurisdictional questions with no life outside that job.
The seven-judge bench addressed the point at paragraph 106 of In Re Interplay, and leaned away from the narrow reading without quite closing the question. It acknowledged that “a plain reading may suggest that Section 16 has incorporated the separability presumption only for the particular purpose of allocation of competence over jurisdictional disputes”. It then recorded the view of the commentator Gary Born that “the better view” is that the presumption states a general rule of contractual validity “which is applicable for all purposes”, and added that “the judicial view that emerges from the Indian courts also seems to suggest that an arbitration agreement is treated as distinct and separate from the underlying contract as a general rule of substantive validity”.
That language repays a close reading, because it is tentative. The bench adopted an academic view and described the Indian case law as seeming to point the same way, which is not the same as laying down a rule. What can fairly be taken from the passage is a direction of travel: the better view treats separability as a rule of substantive validity in its own right, rather than as a device that exists only to feed kompetenz-kompetenz. Paragraph 104 points the same way, recording the UNCITRAL Working Group’s view that the separability presumption sits in Article 16(1) “to complement the principle of competence-competence”. Each doctrine makes the other workable, and neither is the servant of the other.
Positive and negative effects
The positive and negative effects are the two directions in which kompetenz-kompetenz operates: the positive effect speaks to the tribunal and empowers it, and the negative effect speaks to the courts and restrains them. In Re Interplay set out both, and the division has organised every subsequent judgment on the topic.
The positive effect: the tribunal goes first
The positive aspect is the tribunal’s own authority to decide. Chandrachud CJI explained the policy at paragraph 128: arbitration law prioritises the tribunal “by permitting them to initially decide challenges to their authority instead of the courts”, for two reasons. The first is “to recognize the mutual intention of the parties of choosing the arbitrator to resolve all their disputes about the substantive rights and obligations arising out of contract”. The second is “to prevent parties from initiating parallel proceedings before courts and delaying the arbitral process”.
The second reason is the practical one. A jurisdictional objection is the cheapest available delaying tactic, and a regime that routed every such objection to a court first would make the tactic free.
The negative effect: the courts stand back
The negative aspect is the instruction to courts. At paragraph 129 the Court held that it “instructs the courts to limit their interference at the referral stage by deferring to the jurisdiction of the arbitral tribunal in issues pertaining to the existence and validity of an arbitration agreement”, so that courts “should refrain from entertaining challenge to the jurisdiction of the arbitral tribunal before the arbitrators themselves have had an opportunity to do so”. Folding both facets together, the bench defined the principle as “a rule whereby arbitrators must have the first opportunity to hear challenges relating to their jurisdiction, which is subject to subsequent review by courts”.
Vidya Drolia had already drawn a distinction inside the negative effect that still matters. At paragraph 129 of the report, Khanna J separated the expressed negative effect, where a statute itself confines the court and the court simply follows it, from the implied negative effect, which “curtails and constrains interference by the court at the referral stage by necessary implication in order to allow the arbitral tribunal to rule as to their jurisdiction”. Section 11(6A) is the expressed version. Most of the rest is implied.
The negative effect binds constitutional courts too. In Bhaven Construction v. Executive Engineer, Sardar Sarovar Narmada Nigam Ltd., 2021 SCC OnLine SC 8, decided on 6 January 2021 and reported at (2022) 1 SCC 75, a party whose Section 16 plea had been rejected went to the High Court under Articles 226 and 227 rather than waiting for the award. Ramana J held at paragraph 17 that the power to interfere “needs to be exercised in exceptional rarity, wherein one party is left remediless under the statute or a clear ‘bad faith’ shown by one of the parties”, and at paragraph 25 that Section 16 “necessarily mandates that the issue of jurisdiction must be dealt first by the tribunal, before the Court examines the same under Section 34”. A losing party on a jurisdictional plea is not remediless. It is early.
What the referral court may still examine
The referral court may examine only the prima facie existence of the arbitration agreement, and Section 11(6A) says so in terms: the court “shall, notwithstanding any judgment, decree or order of any court, confine to the examination of the existence of an arbitration agreement”. Everything past existence belongs to the tribunal. The contested part of this topic over the last decade has been how far “existence” was allowed to stretch, and the trajectory runs in one direction.
From SBP to Section 11(6A)
SBP & Co. opened the inquiry in 2005 by holding the Chief Justice’s power under Section 11(6) to be judicial rather than administrative, which let the referral court decide questions the tribunal would otherwise have taken. Parliament closed it. The 2015 amendment inserted Section 11(6A) with a non obstante clause aimed squarely at the case law.
The purpose was stated in M/s Uttarakhand Purv Sainik Kalyan Nigam Ltd. v. Northern Coal Field Ltd., (2020) 2 SCC 455, decided on 27 November 2019. Indu Malhotra J recorded that Section 11 was amended “to overcome the effect of all previous judgments rendered on the scope of power by a non obstante clause, and to reinforce the kompetenz-kompetenz principle enshrined in Section 16”, and held that the court is “now required only to examine the existence of the arbitration agreement. All other preliminary or threshold issues are left to be decided by the arbitrator under Section 16, which enshrines the Kompetenz-Kompetenz principle.” Applying that, the Court sent the question of limitation to the tribunal.
The eye of the needle, and its rejection
The narrowing continued in NTPC Ltd. v. M/s SPML Infra Ltd., 2023 SCC OnLine SC 389, decided on 10 April 2023, which gave the doctrine its best-known label. Under the heading “Eye of the Needle”, Narasimha J held that the pre-referral jurisdiction under Section 11(6) “is very narrow and inheres two inquiries”, a primary inquiry into existence and validity and a secondary one into non-arbitrability, and at paragraph 28 that “the limited scrutiny, through the eye of the needle, is necessary and compelling”. The threshold was deliberately severe: refer unless there is “not even a vestige of doubt” that the claim is non-arbitrable.
Then the Supreme Court went further than its own test. Krish Spinning held in July 2024 that the eye of the needle is itself too wide. Pardiwala J reasoned at paragraph 118 that tests of that kind, “although try to minimise the extent of judicial interference, yet they require the referral court to examine contested facts and appreciate prima facie evidence (however limited the scope of enquiry may be) and thus are not in conformity with the principles of modern arbitration which place arbitral autonomy and judicial non-interference on the highest pedestal”.
Paragraph 116 states the consequence for the negative effect in unusually strong terms: a matter falling within the tribunal’s exclusive domain “should not be looked into by the referral court, even for a prima facie determination, before the arbitral tribunal first has had the opportunity of looking into it”. The objection to the older tests is not that they let courts decide too much. It is that any factual appreciation, however light, is a form of deciding.
“Examine” and “rule” are different verbs
The textual argument Krish Spinning built this on repays attention, because it is available to any practitioner arguing a Section 11 petition. At paragraph 111 the Court read the two provisions against each other:
“The use of the term ‘examination’ under Section 11(6-A) as distinguished from the use of the term ‘rule’ under Section 16 implies that the scope of enquiry under section 11(6-A) is limited to a prima facie scrutiny of the existence of the arbitration agreement, and does not include a contested or laborious enquiry, which is left for the arbitral tribunal to ‘rule’ under Section 16.”
Parliament used two different verbs, and the difference between them carries the allocation. Courts examine. Tribunals rule.
The position as it stands in 2026 was restated in Motilal Oswal Financial Services Ltd. v. Santosh Cordeiro, 2026 INSC 5, decided on 5 January 2026. Viswanathan J put it at paragraph 17: “our jurisdiction is only to inspect or scrutinize the dealings between the parties for determination about the existence of an Arbitration Agreement. We are not to launch a laborious or a contested inquiry.”
Two boundaries frame this section. Non-arbitrability is a separate gate with its own case law, set out in our guide to the arbitrability of disputes in India. And the appointment machinery the referral court operates, once it is satisfied an agreement exists, is covered in LawSikho’s treatment of Section 11 and the appointment of an arbitrator.
Who decides first: the referral court’s lane and the tribunal’s lane
Referral court, Section 11(6A)
EXAMINE
- Prima facie existence of the arbitration agreement
- Formal validity: is it in writing under Section 7
- Nothing that needs contested evidence
- No “laborious or contested inquiry” (Motilal Oswal, 2026)
- Not even a prima facie look at the tribunal’s exclusive domain (Krish Spinning, 2024)
Arbitral tribunal, Section 16
RULE
- Existence and validity of the arbitration agreement
- Who is a party, including non-signatories
- Scope of the disputes referred
- Limitation (Uttarakhand Purv Sainik, 2019)
- Accord and satisfaction, discharge vouchers, duress (Krish Spinning, 2024)
The textual hook: Parliament used two different verbs, and the difference carries the allocation. Courts examine. Tribunals rule. (Krish Spinning, 2024 INSC 532, para 111)
How the court’s lane narrowed
- 2005 SBP: referral court’s inquiry opened
- 2015 Section 11(6A) inserted: confined to existence
- 2019 Uttarakhand Purv Sainik: limitation to the tribunal
- 2023 NTPC: the “eye of the needle”
- 2023 In Re Interplay (7 judges): prima facie existence only
- 2024 Krish Spinning: even the eye of the needle is too wide
Sources: Arbitration and Conciliation Act, 1996, ss. 11(6A) and 16; Krish Spinning (2024 INSC 532); In Re Interplay (2023 INSC 1066); Motilal Oswal (2026 INSC 5)iPleaders
Where separability stops
Separability stops where the defect is aimed at the arbitration agreement itself rather than at the contract around it. The doctrine is a shield against contamination from the host contract, not a rule that arbitration clauses are indestructible. In Re Interplay recorded the qualification early: discussing the limits of the presumption at paragraph 93, the bench noted that “in certain situations, the defects in the underlying contract could also invalidate the arbitration agreement”, and at paragraph 104, summarising the position under Article 16(1) of the Model Law, that the tribunal’s jurisdiction “will be affected only when the defect causing invalidity is directed at the arbitration agreement”.
The formation test
The question to ask is not whether the contract is void. It is whether the vitiating factor attacks the formation of the arbitration agreement itself.
Khanna J drew the line in his concurring opinion in In Re Interplay. At paragraph 28 he held that invalidity of an underlying agreement “may not, unless relating to its formation, result in invalidity of the arbitration clause in the underlying agreement”. At paragraph 39 he identified the residual category, holding it important “to draw a distinction between contracts which are void, and those which are void ab initio due to lack of elements of offer or acceptance, which prevents a contract from coming into operation”.
That distinction does the whole job. A contract that was formed but is void, illegal, repudiated, discharged or unstamped leaves the clause standing, because the parties did agree to arbitrate and their agreement is independent. Where no agreement was formed at all, there is no clause either, because separability presupposes that something was made. The oldest statement of the point is still the sharpest: Lord Macmillan in Heyman, at page 371, held that “if there has never been a contract at all, there has never been as part of it an agreement to arbitrate. The greater includes the less.”
Forgery and the clause that was never agreed
The Supreme Court applied the limit in Rajia Begum v. Barnali Mukherjee, 2026 INSC 106, decided on 2 February 2026, where the document said to contain the arbitration clause was alleged to be forged. Aradhe J held at paragraph 20:
“In the present case, arbitration clause does not exist independently but is embedded in the document whose existence is seriously disputed. Arbitration, it bears reiteration, is founded upon consent. A party may be bound by the arbitral process only if it is first shown, even at a prima facie level, that such a party had agreed to submit disputes to arbitration. Where the arbitration agreement itself is alleged to be forged or fabricated, the disputes ceases to be merely contractual and strikes at the very root of arbitral jurisdiction. A controversy of this nature falls squarely within the category of disputes that are generally recognized as non-arbitrable.”
At paragraph 23 the Court drew the practical conclusion: “When the existence of the arbitration agreement itself is in serious dispute and requires adjudication, appointment of an arbitrator would be premature and legally impermissible.”
Three points of precision are owed to that judgment. The route the Court took was non-arbitrability rather than separability: the closing words of paragraph 20 place a forged agreement in the category of disputes “generally recognized as non-arbitrable”, which is the same destination reached by a different road. The judgment does not use the words “separability”, “kompetenz-kompetenz” or “Section 16” anywhere, so its significance for these doctrines lies in what it decides rather than in any vocabulary it adopts. And the Court’s language at paragraph 24 confines the ruling to “this stage” rather than declaring the dispute permanently beyond arbitration.
English law reaches the same place by the same reasoning, which is why Indian courts cite it. Lord Hoffmann gave the example in Premium Nafta at paragraph 17: where the main agreement and the arbitration agreement sit in one document “and one of the parties claims that he never agreed to anything in the document and that his signature was forged, that will be an attack on the validity of the arbitration agreement”. Lord Hope set the standard for getting there: “The doctrine of separability requires direct impeachment of the arbitration agreement before it can be set aside. This is an exacting test. The argument must be based on facts which are specific to the arbitration agreement. Allegations that are parasitical to a challenge to the validity to the main agreement will not do.”
Fraud follows the same contour. Vidya Drolia overruled the older position that fraud allegations were inherently non-arbitrable, subject to a caveat Khanna J recorded at paragraph 78 of the report: the exception is “fraud, which would vitiate and invalidate the arbitration clause”. Fraud in the contract goes to the tribunal. Fraud in the clause does not.
Novation: an open question
Novation is the one area where the position is genuinely unsettled, and it deserves to be reported as unsettled rather than tidied up.
The traditional rule came from Union of India v. Kishorilal Gupta & Bros, 1959 AIR 1362: where parties replace a contract with a new one, the arbitration clause in the superseded contract perishes with it. In Re Interplay addressed those older authorities at paragraph 108, holding that Section 16(1)(b) “renders the decisions in Kishorilal Gupta and Damodar Valley Corporation redundant”, so that “even if the underlying contract is declared null and void, it will not ipso jure result in the invalidity of the arbitration agreement”.
Read that carefully before concluding the novation rule is gone. The bench said “redundant”, not “overruled”, and its reason was Section 16(1)(b), which addresses a contract being null and void. Novation is not nullity: a novated contract was valid and was replaced. On the current authorities the substantive rule about novation appears to survive, and what has clearly changed is who decides it. In Sanjiv Prakash v. Seema Kukreja, 2021 SCC OnLine SC 282, the Supreme Court declined to decide whether a memorandum had been novated, holding at paragraph 9 that the question required detailed consideration of both agreements and their surrounding circumstances, and that “none of this can be done given the limited jurisdiction of a court under Section 11”. The question went to the tribunal.
That reading is analysis rather than holding, and it is offered as such. No judgment has expressly overruled the line of authority applying the novation rule under the 1996 Act, and none has expressly preserved it. What a practitioner can say with confidence today is that the argument is available and that the forum for it is the tribunal.
Where separability stops: what is the attack aimed at?
What is the attack aimed at?
The main contract
- Contract is null and void (s. 16(1)(b))
- Repudiation or breach (Heyman, 1942)
- Discharged by accord and satisfaction (Krish Spinning, 2024)
- Unstamped: a curable defect (In Re Interplay, 2023)
- Illegal or voidable
- Fraud in the contract (Vidya Drolia, para 78)
Clause SURVIVES. The tribunal rules under Section 16.
The arbitration clause itself
- Signature on the document alleged forged (Rajia Begum, 2026)
- Party says it never agreed to anything in the document (Premium Nafta, 2007)
- No offer or acceptance, so nothing was formed (In Re Interplay, Khanna J, para 39)
- Fraud that vitiates the clause itself (Vidya Drolia, para 78)
Clause does NOT survive. Existence is decided first; appointment would be premature.
The test: not whether the contract is void, but whether the defect is directed at the formation of the arbitration agreement itself. Invalidity of the underlying agreement “may not, unless relating to its formation, result in invalidity of the arbitration clause”. (In Re Interplay, 2023 INSC 1066, Khanna J concurring, para 28)
Sources: In Re Interplay (2023 INSC 1066); Rajia Begum (2026 INSC 106); Krish Spinning (2024 INSC 532); Premium Nafta [2007] UKHL 40; Heyman [1942] AC 356iPleaders
Where Indian law departs from the Model Law
Indian law departs from the Model Law by leaving out its principal check on the tribunal, which makes Section 16 stronger than the provision it was modelled on. Section 16 “is based on Article 16 of the Model Law”, as In Re Interplay recorded at paragraph 120, and recognises competence-competence in Indian arbitration law. The borrowing was close but not complete, and the omission is the interesting part.
The thirty-day review India left out
Article 16(3) of the Model Law gives a party a fast route to court. Where a tribunal rules as a preliminary question that it has jurisdiction, any party may ask the designated court, “within thirty days after having received notice of that ruling”, to decide the matter; the court’s decision is subject to no appeal, and the tribunal may continue meanwhile. UNCITRAL’s own explanatory note describes that combination as the safeguard against dilatory tactics.
India did not enact Article 16(3). Section 16(5) and 16(6) put something else in its place: where the tribunal rejects a jurisdictional plea it continues and makes an award, and the aggrieved party challenges that award under Section 34. There is no thirty-day interlocutory run to court.
The consequence is structural rather than technical. Indian kompetenz-kompetenz bites harder than the Model Law baseline because the Model Law’s mid-arbitration escape hatch does not exist here. It also explains Bhaven Construction: a party denied a Section 16 plea was seeking, through writ jurisdiction, something close to the Article 16(3) review that Parliament chose not to enact. Read that way, the “exceptional rarity” standard is not judicial reticence but fidelity to a deliberate legislative choice.
Section 11(6A): deleted, un-notified, still law
The most striking feature of this area is an anomaly. Section 11(6A), the provision that confines the referral court to examining existence, was omitted by the Arbitration and Conciliation (Amendment) Act, 2019, Act 33 of 2019. That omission has never been brought into force. The provision Parliament deleted is still law, because the deletion was never notified.
Courts have not overlooked this. It has been confirmed in an unbroken judicial line. In Re Interplay held at paragraph 152 that Vidya Drolia’s contrary premise was “erroneous because the omission of Section 11(6A) has not been notified and, therefore, the said provision continues to remain in full force”. Cox and Kings Ltd. v. SAP India Pvt. Ltd., 2023 INSC 1051, a Constitution Bench, recorded the same at paragraph 156. So did Ajay Madhusudan Patel v. Jyotrindra S. Patel, 2024 INSC 710 at paragraph 56. Most recently, Motilal Oswal noted at paragraph 14 that Section 11(6A), “though omitted by Act 33 of 2019, the omission has not yet been notified”.
The wider legislative picture is similarly suspended, and any statement of it needs a date. As at July 2026, the Arbitration and Conciliation Act, 1996 remains the operative statute. The draft Arbitration and Conciliation (Amendment) Bill circulated for public consultation in October 2024, on the recommendations of an expert committee, has not been enacted. Speaking on 11 July 2026, the Chief Justice of India said that the draft Bill “remains exactly that, and its new avatar is hopefully now being introduced in Parliament”, adding that “if our ambition is to become a preferred seat, this gap between announcement and implementation is precisely the credibility deficit we cannot legislate our way out of”. He also noted that the Arbitration Council of India, created on the statute book by the 2019 amendment, has still not been constituted. That last phrase leaves the position live, so readers coming to this later should check whether the Bill has since been introduced.
For anyone drafting across borders, the practical lesson is that choosing a seat means choosing a jurisdictional regime and not merely a hearing room, which is worth holding onto when drafting contracts for foreign clients. The same clause, seated in a Model Law jurisdiction that enacted Article 16(3), gives a party a mid-arbitration route to court that the identical clause seated in India does not.
Frequently asked questions
What is the difference between kompetenz-kompetenz and separability?
Kompetenz-kompetenz decides who rules on a jurisdictional objection and in what order, giving the arbitral tribunal the first opportunity. Separability decides what the arbitration clause is, treating it as an agreement independent of the contract containing it. The first allocates authority between tribunal and court; the second insulates the clause from defects in its host contract.
Does the arbitration clause survive if the main contract is void?
Yes. Section 16(1)(b) of the Arbitration and Conciliation Act, 1996 provides that a tribunal’s decision that the contract is null and void “shall not entail ipso jure the invalidity of the arbitration clause”. In Re Interplay confirmed at paragraph 108 that even a contract declared null and void does not automatically invalidate the arbitration agreement. The clause also survives repudiation, breach and discharge by settlement.
Can a court decide whether an arbitration agreement exists, or must the tribunal?
Both can, at different stages and to different depths. Under Section 11(6A), a referral court examines the prima facie existence of the agreement and nothing more. Krish Spinning held in 2024 that the court must not conduct a contested or laborious inquiry, because Section 11(6A) says “examine” while Section 16 says “rule”. The tribunal then rules on existence and validity fully, subject to review under Section 34.
What happens if the arbitration clause itself is alleged to be forged?
Separability does not save it. In Rajia Begum v. Barnali Mukherjee, decided in February 2026, the Supreme Court held that where the arbitration agreement itself is alleged to be forged or fabricated, the dispute “strikes at the very root of arbitral jurisdiction”, and that appointing an arbitrator would be “premature and legally impermissible”. The test is whether the attack is directed at the formation of the arbitration agreement rather than at the contract around it.
Is Section 11(6A) still in force after the 2019 amendment?
Yes. Act 33 of 2019 omitted Section 11(6A), but that omission has never been notified into force, so the provision continues to operate. The Supreme Court has confirmed this repeatedly, in In Re Interplay at paragraph 152, Cox and Kings at paragraph 156, Ajay Madhusudan Patel at paragraph 56, and most recently Motilal Oswal in January 2026.
References
Case law
- A. Ayyasamy v. A. Paramasivam, (2016) 10 SCC 386
- Adavya Projects Pvt. Ltd. v. M/s Vishal Structurals Pvt. Ltd., 2025 INSC 507
- Ajay Madhusudan Patel v. Jyotrindra S. Patel, 2024 INSC 710
- Arabian Exports Private Ltd. v. National Insurance Co. Ltd., 2025 INSC 630
- Bhaven Construction v. Executive Engineer, Sardar Sarovar Narmada Nigam Ltd., 2021 SCC OnLine SC 8, reported at (2022) 1 SCC 75
- Cox and Kings Ltd. v. SAP India Pvt. Ltd., 2023 INSC 1051, 2023 SCC OnLine SC 1634 (Constitution Bench)
- Enercon (India) Ltd. v. Enercon GmbH, (2014) 5 SCC 1, AIR 2014 SC 3152
- Heyman v. Darwins Ltd, [1942] AC 356 (House of Lords)
- In Re: Interplay Between Arbitration Agreements under the Arbitration and Conciliation Act 1996 and the Indian Stamp Act 1899, 2023 INSC 1066 (seven-judge bench)
- Motilal Oswal Financial Services Ltd. v. Santosh Cordeiro, 2026 INSC 5
- NTPC Ltd. v. M/s SPML Infra Ltd., 2023 SCC OnLine SC 389
- Premium Nafta Products Ltd. v. Fili Shipping Co. Ltd., [2007] UKHL 40, on appeal from Fiona Trust & Holding Corp v. Privalov, [2007] EWCA Civ 20
- Prima Paint Corp. v. Flood & Conklin Mfg. Co., 388 U.S. 395 (1967) (Supreme Court of the United States)
- Rajia Begum v. Barnali Mukherjee, 2026 INSC 106
- S.B.P. & Co. v. Patel Engineering Ltd., (2005) 8 SCC 618 (seven-judge bench)
- Sanjiv Prakash v. Seema Kukreja, 2021 SCC OnLine SC 282
- SBI General Insurance Co. Ltd. v. Krish Spinning, 2024 INSC 532, 2024 SCC OnLine SC 1754
- Union of India v. Kishorilal Gupta & Bros, 1959 AIR 1362, 1960 SCR (1) 493
- M/s Uttarakhand Purv Sainik Kalyan Nigam Ltd. v. Northern Coal Field Ltd., (2020) 2 SCC 455
- Vidya Drolia v. Durga Trading Corporation, (2021) 2 SCC 1
Statutes and instruments
- Arbitration and Conciliation Act, 1996, sections cited: 7, 11(6), 11(6A), 16, 34
- Arbitration and Conciliation (Amendment) Act, 2015 (Act 3 of 2016), inserting Section 11(6A)
- Arbitration and Conciliation (Amendment) Act, 2019 (Act 33 of 2019), omitting Section 11(6A) (not notified)
- UNCITRAL Model Law on International Commercial Arbitration, 1985, with amendments as adopted in 2006, Article 16
Other sources
- CJI flags delay in creation of Arbitration Council of India, Outlook India (PTI), 11 July 2026, reporting the Chief Justice of India’s remarks at the Indian Institute of Arbitration and Mediation silver jubilee
This article is for informational and educational purposes only and does not constitute legal advice. Readers should consult a qualified advocate for advice on their specific circumstances.
