King, Stubb & Kasiva | View firm profile
The validity of an arbitration agreement under Section 7 of the Arbitration and Conciliation Act, 1996 (Act) ultimately boils down to consent. At its core, an arbitration clause is nothing more than a contractual promise by the parties to resolve their disputes outside the regular court system. But consent isn’t always straightforward. Sometimes parties sign a contract containing a clause that is too weak or tentative to create any real obligation to arbitrate. In other situations, a party may perform extensively under a contract it never signed, leaving the question open whether it is bound by the arbitration clause contained in it. These are two quite different problems, and both come up fairly often in practice.
The Hon’ble Supreme Court has addressed these two pertinent issues in two distinct judgments which together mark the outer scope of what Section 7 of the Act requires. In Nagreeka Indcon Products Pvt. Ltd. v. Cargocare Logistics (India) Pvt. Ltd. (2026 INSC 384) (Nagreeka), the Court held that a clause providing that disputes “can” be settled by arbitration does not create a binding arbitration agreement. On the other hand, in Glencore International AG v. Shree Ganesh Metals and Another (2025 INSC 1036) (Glencore), the Court ruled that an unsigned but clearly mandatory arbitration clause is enforceable if the non-signing party’s conduct shows unequivocal acceptance of the contract. When these two decisions are read together, a clear principle on the scope of Section 7 is established that conduct can fill the gap left by an unsigned agreement, but it cannot fix a clause that, by its own wording, does not create a binding obligation on the parties.
I. WHEN THE CLAUSE ITSELF IS NOT ENOUGH: NAGREEKA
While Section 7 of the Act requires an arbitration agreement to be in writing, it does not specify what language is sufficient to create one. That question has been settled by a consistent line of decisions of the Supreme Court. The Court had earlier identified the essential attributes of a valid arbitration agreement in K.K. Modi v. K.N. Modi (1998) 3 SCC 573), which required, inter alia, that the agreement contemplate a binding decision by the tribunal deriving from the consent of the parties, and that the agreement to refer disputes to the tribunal be intended to be enforceable in law.
More recently, the Supreme Court in Jagdish Chander v. Ramesh Chander ((2007) 5 SCC 719) (Jagdish Chander), has authoritatively held that the words of an arbitration clause must disclose a determination and obligation to go to arbitration, not merely the possibility of doing so. A clause that provides that parties “can, if they so desire” refer disputes to arbitration is not an arbitration agreement; it is an agreement to consider entering into one, contingent on fresh consent when a dispute actually arises. It has also made clear that mere use of the word “arbitration” or “arbitrator” will not make a clause a valid arbitration agreement if it requires or contemplates fresh consent at the time of a dispute.
The decision in Nagreeka is a direct application of this principle laid down by the Supreme Court in Jagdish Chander. The dispute arose from a bill of lading wherein the Clause 25 provided that differences between the parties “can be settled by arbitration in India or a place mutually agreed with each party appointing an arbitrator.” Subsequent thereto, when a payment dispute arose, the appellant invoked this clause and sought appointment of an arbitrator under Section 11 of the Act. when a dispute with respect to payments arose and the appellant sought appointment of an arbitrator under Section 11 of the Act. The respondent, however, refused to participate in arbitration, contending that the clause imposed no binding obligation to do so. Both the Bombay High Court and the Supreme Court agreed. The Bombay High Court dismissed the Section 11 application, and the Supreme Court upheld that dismissal.
The Court’s reasoning was based on a careful reading of the word “can“. Examining its ordinary dictionary meaning across multiple sources, the Court held that “can” denotes capacity or factual possibility, not obligation. It contrasted the word “can” with “shall,” which signals a mandate, and noted that even “may” carries a stronger obligatory weight in judicial interpretation than “can” does. The Court then applied this understanding to Clause 25 and held that the clause indicated merely the future possibility of referring disputes to arbitration. For disputes to actually be settled by arbitration, a further agreement between the parties would be required, and such an agreement could only come into existence if both parties consented to it. Since the respondent had refused to refer the matter to arbitration, no such further agreement existed.
The Court also dealt with several arguments advanced by the Appellants. The first argument was that the heading of the clause read “Arbitration,” which the Appellant argued demonstrated the parties’ intention. The Court rejected this, holding that a heading cannot supply a mandatory obligation that the body of the clause does not contain. The second argument was that the Section 11 stage requires only a prima facie examination of whether an arbitration agreement exists, and that doubts should be resolved in favour of referral. The Court clearly dealt with this argument by placing reliance on SBI General Insurance Co. Ltd. v. Krish Spg. ((2024) 12 SCC 1), it acknowledged this limitation but held that it did not assist the appellant, because the absence of a binding arbitration clause was “manifestly and ex facie certain” on the face of Clause 25 itself. The prima facie threshold does not require courts to refer parties to arbitration where the non-existence of the arbitration agreement is plain on the document. This reading aligns with the caution expressed in Goqii Technologies (P) Ltd. v. Sokrati Technologies (P) Ltd. ((2025) 2 SCC 192), that the limited jurisdiction of referral courts must not be misused to compel participation in arbitration on the basis of non-existent agreements.
The significance of Nagreeka:
Nagreeka adds to the existing line of judgments in two respects. First, it extends the mandatory obligation analysis to the word “can,” in the arbitration context. The Court’s close textual analysis of ordinary dictionary usage and its comparison with “shall” and “may” provides a useful linguistic framework for evaluating dispute resolution clauses going forward. Second, the decision confirms that no combination of contextual factors will cure a fundamentally permissive clause. The presence of the word “Arbitration” as a heading, the commercial context of the transaction, and the general judicial preference for referring commercial disputes to arbitration were all considered and all found insufficient. The clause itself is dispositive. What makes Nagreeka useful for practitioners is its specificity. A clause using “can” or “may” will ordinarily be treated as permissive and, absent a clear contextual correction, will not constitute a valid arbitration agreement.
II. WHEN THE SIGNATURE IS MISSING BUT THE CONSENT IS NOT: GLENCORE
The second problem is structurally the inverse of the first. Where Nagreeka concerned a clause whose language was deficient, Glencore concerned a clause whose language was unambiguously mandatory but which had not been signed by one of the parties. The question was whether extensive performance under the contract could supply the missing assent.
Section 7(3) of the Act requires an arbitration agreement to be in writing. It does not specifically require a signature. Moreover, Section 7(4)(b) of the Act expressly provides that an exchange of communications providing a record of the agreement is sufficient. This much has been settled since Jugal Kishore Rameshwardas v. Goolbai Hormusji (AIR 1955 SC 812), which established that formal execution is not a prerequisite for a valid arbitration agreement.
The essential ingredients of a valid arbitration agreement as identified in Bihar State Mineral Development Corporation v. Encon Builders (I) (P) Ltd. ((2003) 7 SCC 418) include, relevantly, that the parties must intend to settle their disputes by a private tribunal and must agree in writing to be bound by its decision. The requirement of intention and agreement in writing does not, however, require that the writing take the specific form of a document signed by all parties.
Section 7(4)(b) expressly recognises that an exchange of communications establishing a record of the agreement is sufficient. The principle was further developed in Great Offshore Ltd. v. Iranian Offshore Engineering and Construction Co. ((2008) 14 SCC 240), where the Court held that procedural formalities such as signatures, stamps, seals are red tape and not conclusive if the parties can demonstrate their intention to arbitrate through other justiciable means.
Glencore applies this framework in a factual scenario where the conduct evidence was exceptionally strong. The Respondent, Shree Ganesh Metals had never signed the contract containing the arbitration clause, which designated London as the seat. However, it accepted delivery of zinc metal against invoices that each specifically referenced the contract by its unique number. It instructed its bank to issue standby letters of credit that also expressly referenced the same contract. And it sent a direct written communication to Glencore confirming that it would not default in performance under the contract.
The Bombay High Court nonetheless refused to refer the parties to arbitration, holding that no concluded contract existed in the absence of a signature. The Supreme Court reversed the concurrent findings of the Single Judge and Division Bench of the Bombay High Court, holding that the court had failed to give proper weightage to the conduct evidence. It held that Shree Ganesh’s consistent engagement with the specific contract document demonstrated unequivocal acceptance of its terms, including the arbitration clause. The Court confirmed that signatures are not required under Sections 7(4)(b), 7(4)(c), or 7(5) of the Act, and directed referral to arbitration under Section 45.
It is to be considered that the conduct evidence in Glencore was notable for being referential rather than merely consistent. Shree Ganesh did not simply perform obligations that the contract, it did so against documentation that specifically referred the underlying contract by its reference number across multiple independent transactions. This is a pertinent and a material distinction. Conduct that is merely consistent with a contract (for example, accepting delivery of goods without specifically acknowledging the underlying contract document) raises a different evidentiary question from conduct that specifically references the contract and thereby demonstrates that the non-signing party has engaged with its terms. The strength of Glencore as a precedent lies precisely in this specificity, and courts applying the conduct-based framework in future cases will need to attend to the quality of the conduct evidence, not merely its existence.
The Significance of Glencore:
Glencore clearly settles two pertinent questions that frequently arise. First, it confirms that a non-signing party which has specifically engaged with a contract document across multiple independent transactions cannot resist arbitration on the basis of non-signature alone. The Court’s emphasis on referential conduct draws a workable evidentiary line, one where documentation that identifies the contract by name or number carries greater probative weight than performance that is merely consistent with the contract’s existence. Second, the decision reinforces that the conduct-based inquiry under Section 7(4)(b) is not a low threshold. A party seeking to establish an arbitration agreement through conduct must demonstrate that the other party engaged with the specific contract, not merely that it behaved in a manner the contract contemplated.
For practitioners, the practical lesson cuts in both directions. A party relying on an unsigned contract should ensure that all downstream documentation, including invoices, letters of credit, and written communications, specifically references the contract by its identifier. Conversely, a party that has performed under a contract it did not intend to be bound by must raise that objection early and clearly, since consistent performance against specific contractual references will be treated as unequivocal acceptance of all the contract’s terms, including its arbitration clause.
III. THE COMMON PRINCIPLE BETWEEN NAGREEKA AND GLENCORE:
Nagreeka and Glencore address different facets of failures of consent, but the principle underlying both decisions is the same. An arbitration agreement requires genuine consent to arbitrate, and courts will look to the substance of the parties’ relationship rather than the form of their documentation to determine whether that consent exists.
In the mandatory obligation line, the inquiry is whether the parties’ chosen language reflects a present agreement to be bound by arbitration, or merely a tentative arrangement to consider it. Where the clause is permissive, no amount of subsequent conduct can transform it into a binding agreement. The parties have, by their own words, reserved the right to choose arbitration afresh when a dispute arises. That reservation cannot be overridden by a court.
In the conduct and communications line, the inquiry is whether the parties’ dealings, taken as a whole, reflect a genuine agreement to arbitrate. Where the arbitration clause is itself mandatory, the absence of a signature is a problem of form rather than substance, and conduct that clearly demonstrates acceptance of the clause will cure it, particularly where that conduct specifically references the contract rather than merely being consistent with its existence.
The order in which courts approach these questions also matters. A court must first satisfy itself that the clause in question actually imposes an obligation to arbitrate. If it does not, the question of whether a party is bound through conduct simply does not arise. Nagreeka confirms that this threshold inquiry remains available even at the Section 11 stage. A clause that is plainly permissive on its face can be identified and rejected at that stage itself, without sending the parties through the time and expense of constituting an arbitral tribunal.
IV. CONCLUSION:
Section 7 of the Act ultimately turns on whether the parties genuinely agreed to resolve their disputes by arbitration. The form of that agreement is quite flexible. It can take the shape of a signed document, an exchange of communications, or clear conduct showing acceptance of a mandatory arbitration clause. What it cannot be is a permissive clause that merely leaves the option of arbitration open for a future decision, or conduct that tries to create an obligation which the parties’ own wording did not establish.
When read together, Nagreeka and Glencore mark the two ends of the spectrum. For practitioners, the practical takeaway is clear. The strength of any dispute resolution clause depends on the obligation it actually creates. A clause using “can” or “may” does not amount to a binding arbitration agreement. It is only an invitation that either party can decline. By contrast, when the clause uses clear mandatory language such as “shall”, a party who has performed under the contract cannot easily escape arbitration merely by pointing to the absence of its signature. In the end, there are no shortcuts. Careful drafting that leaves no room for ambiguity, along with proper documentation of performance and acceptance, remains the safest safeguard on both sides.
AUTHORED BY – PRAGALBH BHARDWAJ, ASSOCIATE PARTNER,
KING STUBB AND KASIVA