THE SUPREME COURT’S Court’s Gayatri Balasamy v. ISG Novasoft Technologies Ltd., (2025) decision was heralded as a watershed on the scope of Sections 34 and 37 of the Arbitration Act. A 4:1 majority held that courts have a limited power to modify arbitral awards in the interests of justice. On its face, this reading appears more pragmatic than earlier case law, aiming to avoid endless re-arbitration on technical grounds. But a closer look reveals that even this “progressive” stance fails to stem excessive judicial intervention.
In effect, the judgment reopens the door to judicial rewriting of awards, undermining party autonomy and delaying finality. The majority’s interpretation of Sections 34 and 37 still falls far short of the non-interventionist ideal embodied in the UNCITRAL Model Law and reflected in leading arbitration jurisdictions. In particular, the Court’s reasoning on modification, invoking broad principles like actus curiae neminem gravabit and even its plenary Article 142 powers, threatens to dilute the clear legislative scheme and invite further litigation. This was a missed opportunity: in Project Director, NHAI v. M. Hakeem (2021), a Supreme Court bench had emphatically rejected any general modification power, a rule that the 2025 majority in Balasamy could have reaffirmed but instead only partially preserved.
This piece argues that Gayatri Balasamy ultimately fails to create a robust, non-interventionist arbitration regime in India. It sketches the majority’s key holdings but shows how they perpetuate judicial overreach. I further emphasise that sanctioning even “limited” modifications (including correcting interest, severing parts of awards, or applying Article 142) undermines the very principles of party autonomy and finality that arbitration rests upon. The decision’s practical effect is to retain (and even expand) the lengthy Section 34/37 appellate labyrinth, rather than eliminating it. In contrast, common-law jurisdictions like England, Singapore and Australia hew closely to the Model Law’s command that “no court shall intervene” except by express provision.
By contrast, the Indian majority’s approach leaves parties and tribunals exposed to continuing judicial second-guessing. I conclude that the judgment, while attempting to balance efficiency and justice, risks entrenching judicial activism. The real solution lies in clear legislative reform, not incremental judicial expansion of authority.
The decision’s practical effect is to retain (and even expand) the lengthy Section 34/37 appellate labyrinth, rather than eliminating it.
The Majority’s stance on Sections 34 and 37
The Constitution Bench in Gayatri Balasamy confronted the central question: Does the High Court’s power to set aside an arbitral award (Section 34) inherently include a narrower power to “modify” it?
A 4:1 majority answered “yes, in limited circumstances”, but only where doing so does not contradict Section 34’s text. In essence, the Court held that Section 34’s remedy of annulment should be read as coextensive with a “limited power to modify the award within the confines of Section 34”. Importantly, the majority confined this power to “computational, clerical or typographical errors” and other manifest defects that arise on the face of the award. For example, a tribunal’s mis-computation of interest could be corrected by the court (the majority drew explicitly on the ancient maxim actus curiae neminem gravabit). The Court also endorsed severing any “invalid” (non-arbitrable) portions of an award from the valid ones (relying on separability). In a controversial move, it even recognised that a court might invoke its plenary Article 142 power (to do “complete justice”) in very rare cases. Notably, the majority reaffirmed that “Section 37’s appellate jurisdiction is coextensive” with Section 34, meaning a successful modification by a trial court could, on appeal, be further affirmed or altered by the High Court or Supreme Court under the same limited standards.
The majority justified this limited modification power on practical grounds. It noted that refusing any modification would force parties into a new arbitration just to effect minor corrections, a result “more cumbersome than traditional litigation”. In India, applications under Section 34 and appeals under Section 37 already “often take years to resolve”, so the Court reasoned that allowing a one-stop judicial fix might save the parties from yet another arbitration. On this basis, the Court held that “silence in the Act should not be read as a complete prohibition” on correction. In sum, the majority ruled that subject to strict limits, the judiciary does have certain powers to tweak an award – effectively broadening the judicial toolbox beyond the strict annulment regime the Act originally envisioned.
Judicial intervention persists: Eroding autonomy and finality
This reading, however, represents a half-measure that perpetuates judicial interference rather than curbing it. Even a “limited” power to modify awards inherently empowers courts to rework arbitral decisions. Commentators note that “the essence of a mature arbitration jurisdiction is party autonomy and immunity from courts’ re-evaluation”. By allowing judges to substitute their corrections (even if small) for the tribunal’s determination, the majority undermines both autonomy and finality. Arbitration is fundamentally premised on pacta sunt servanda and finality: parties submit disputes to arbitrators to avoid endless court battles. As one analysis emphasises, “one of the main reasons why parties choose arbitration is the principle of finality, that international arbitral awards cannot be appealed on the merits”. When courts inject themselves by “modifying” outcomes, they weaken this core expectation.
In the majority’s own words, judicial powers under Section 34 are “inherent” and can include limited modifications. In practice, however, these so-called “limited” corrections can easily bleed into merit issues. For instance, adjusting post-award interest (as permitted by the majority) can involve complex factual or legal questions about rates or currency, effectively amounting to an appellate review. Relying on Article 142 to rewrite parts of an award (even if cautious) is truly judicial legislation, far outside the Model Law scheme. The Court’s rationale that the parties implicitly consent to be bound by these implied judicial powers does not change the fact that arbitration agreements are meant to limit court control, not expand it. As one commentator warned, “judicial legislation may create more problems than it would solve,” and any such change “should be legislative and not through ad-hoc judgments”
In short, the majority’s approach transforms Section 34 from a narrow “limited remedy” into a sort of de facto appellate procedure, albeit one cloaked in purpose-language.
Moreover, the majority’s focus on expedience is self-defeating. In theory, correcting an award in Court might save one round of re-arbitration, but it does not end the litigation. Because Section 37 appeals “are co-terminus” with Section 34, every modification by the trial court can be re-litigated on appeal. Parties unhappy with a partial fix will inevitably appeal, prolonging the dispute. The Court itself noted that Section 34/37 processes “often take years” in India. By encouraging judicial tinkering, the judgment may invite additional appeals and uncertainty, rather than expedite closure. In this way, it compromises “finality – a facet of party autonomy”. Finality is not just a slogan; it is the quid pro quo for choosing arbitration over court, as one arbitration scholar has written. When courts continue to second-guess awards through repeated appeals or modifications, they return arbitration toward “traditional litigation” in all but name.
In short, the majority’s approach transforms Section 34 from a narrow “limited remedy” into a sort of de facto appellate procedure, albeit one cloaked in purpose-language. The dissent was correct that “modification of an award involves a judicial substitution of reasoning, or outcome” that truly lies outside the tribunal’s mandate. The majority tries to cabin this by reference to obvious errors and severability, but these concepts are slippery. A tribunal’s decision on interest or severability can be hard to characterize as purely “clerical” or “minor,” and inviting courts to adjust them risks endless debate over what qualifies. Meanwhile, the Act’s express correction mechanism, Section 33, is effectively sidelined. Section 33 entrusts arbitrators (not courts) with correcting errors or adding claims after the award. Yet by allowing courts to do the same things, the majority duplicates and potentially overrides the parties’ original choice of decision-maker. Why should an expert tribunal’s final figures be reset by a judge simply because the judge has a cleaner pen?
At bottom, Gayatri Balasamy maintains a high degree of judicial activism. Even if well-intentioned, it ignores the Arbitration Act’s core message (echoing UNCITRAL Article 5) that “no court shall intervene” beyond the narrow grounds provided. Instead, it embeds gaps into Section 34, forcing arbitrators to wonder if courts will rewrite their decisions. In doing so, it leaves the door open to all the problems the 1996 Act was supposed to cure: delay, cost, and uncertainty.
Section 34 and 37: From limited set-aside to uncertain remedy
Section 34(2) of the 1996 Act, modelled on Article 34 of the UNCITRAL Law, sets out specific, narrowly tailored grounds on which a court “may set aside” an award. The statute conspicuously does not mention any power to vary or remake an award. In fact, the Act provides other remedies (like Section 33 corrections by the tribunal and Section 34(4) remand) for addressing defects. Reading Section 34’s “limited remedy” more broadly, as the majority does, undercuts the textual scheme.
In M. Hakeem v. NHAI (2021), a Constitution Bench had already enforced this strict reading. There, the Court held that under the Act, an award can only be remanded or set aside, not modified under Section 34. In other words, the remedy under 34 is co-terminus with the limited right: if an award is flawed, a party’s only recourse is cancellation or remand, not a judicial re-write. This was a clear statement of legislative intent, grounded in the Act’s alignment with the Model Law’s minimal-intervention philosophy. Gayatri Balasamy purported to uphold this position, stating that the majority “upholds Hakeem” except for a narrow carve-out. But by carving out even “limited” modifications, the majority has effectively diluted Hakeem. Sections 33 and 34(4) already gave tribunals and courts a method to cure errors (by corrections and remands), yet the Court now adds a new “modification” remedy. This shifts outcome-determination power from arbitrators to judges in areas where the Act remains silent.
Section 37 further entrenches the problem. Under Section 37, an appeal lies to a higher court from any Section 34 order (setting aside or refusing to set aside). The majority correctly observed that Section 37’s ambit is “co-terminous” with Section 34. In effect, if a trial court modifies an award (or declines to do so), the same question can be relitigated on appeal. The appellate court “still possesses the power of remand” under Section 34(4) and, by extension, to confirm or alter any modifications. Thus, the judgment authorizes repeated judicial scrutiny: first, the Section 34 court can tweak an award, then the Section 37 court can tweak again. Each stage invites further argument. Far from eliminating intervention, this multiplies it.
The result is clear: even with the majority’s “limited power”, courts are still inserting themselves into awards. Where the 1996 Act envisioned arbitration as a substitute for slow litigation, parties now face two layers of possible challenges. And if an award is remanded under 34(4), a party might still challenge the subsequent outcome. In a country with notoriously long court processes, this keeps the arbitration saga alive. Indeed, the majority’s own commentary lamented that arbitration in India has become “more cumbersome than even traditional litigation” when repeat challenges are forced. Allowing courts to tweak awards therefore undercuts the public policy aim of the Act to finalize disputes swiftly.
M. Hakeem v. NHAI (2021): The missed reset point
The 2021 Hakeem decision was explicitly a chance to draw a bright line. In Hakeem, a five-judge bench held unanimously that the Act does not empower courts to vary arbitral awards – the only powers being remand or annulling them. The Court stressed that “given the limited judicial interference on extremely limited grounds… the ‘limited remedy’ under Section 34 is co-terminous with the ‘limited right’… to set aside or remand”. That ruling became the locus classicus for barring modifications, and was followed by subsequent benches when rejecting award tweaks. Thus, Hakeem reaffirmed the Act’s strict text.
Gayatri Balasamy now treats Hakeem as largely correct, yet it fails to carry forward its essence. While the majority says it “upholds” Hakeem except for error corrections, the practical upshot is that any award can be re-opened on one of four grounds (severability, clerical error, interest, Article 142). In effect, Hakeem’s categorical rejection of modification was diluted into an unwritten exception list. This is the very “legislative” function the GAR analysis warned against: the Court, in effect, amended the Act by judicial fiat.
Had the Court truly taken Hakeem to heart, it would have affirmed that Section 34 allows only what it explicitly says. Instead, the decision treats Hakeem as an almost-black-letter rule except where the majority reads around it. This leaves ambiguity and unpredictability. Justice Viswanathan’s dissent in Gayatri Balasamy lamented that a strictly textual reading as in Hakeem should prevail, and that the majority’s twists amount to usurping Parliament’s role. Indeed, even some prior Supreme Court decisions (like Ssangyong v. NHAI) hinting at using Article 142 for “complete justice” had not been seen as creating a new remedy. By endorsing Article 142 use here, the majority compounds the legislative gap.
Ultimately, Hakeem was a missed opportunity. That case could have decisively closed the door on judicial rewriting of awards. Instead, Gayatri Balasamy sets up a patchwork: courts can modify, but only in theoretically narrow ways that still allow judges to micromanage awards. The result is that Hakeem’s promise of finality has been hollowed out.
True reform lies not in creative jurisprudence but in legislative clarity.
Lessons from Common Law countries
United Kingdom: The Arbitration Act 1996 explicitly embraced the UNCITRAL ethos that arbitration is an alternate forum and parties’ agreement is paramount. Section 1 of the Act declares that “the object of arbitration is to obtain fair resolution of disputes by an impartial tribunal without unnecessary delay or expense,” and that “parties should be free to agree how their disputes are resolved”. Crucially, it adds that “in matters governed by this Part the court should not intervene except as provided by this Part.” In practice, UK courts respect final awards: appeals on questions of law are possible only by leave (and parties can opt out entirely), and courts may only set aside awards on narrow grounds (such as serious irregularity or jurisdiction). The courts also remind themselves that any intervention should “secure the fairness and legitimacy of the system” but not “weaken the process or its finality”. In short, English law codifies a default of non-interference: judges are hands-off unless the Act clearly says otherwise.
Singapore: Under the International Arbitration Act (which implements the Model Law), Singapore courts have likewise adopted a strict non-interventionist posture. There is no right of appeal on a question of law unless the parties have agreed, and in many institutional rules, appeals are disallowed altogether. Singaporean judges have repeatedly emphasized a “policy of minimal intervention” in arbitration. For example, the Court of Appeal in AJU v. AJT explicitly upheld the “autonomous nature of the arbitral process” under a standard of “minimal review”. In every known context, from subpoenas to pre-arbitral discovery, the courts defer to the parties’ chosen procedures unless absolutely necessary, often quoting Article 27 of the Model Law to require tribunal consent before court aid. One Kluwer commentary notes that Singapore courts have “carefully refrained from undue interference in the arbitral process”. Singapore’s experience shows that strict adherence to Article 5 (non-intervention) fosters confidence in arbitration: parties know that, absent agreed exceptions, awards will be final.
Conclusion
In theory, Gayatri Balasamy sought to reconcile the arbitration act’s regime with practical justice. But in substance, it upholds a regime of continued judicial activism. The majority’s allowance of “limited” modifications still undercuts the autonomy and finality that arbitration promises. The judgment actually continues the trend of courts “holding the ring” over Indian arbitration, rather than stepping back. What India needed – and Hakeem hinted at – was a bright line: awards stand unless set aside for valid grounds, not tinkered with. Unfortunately, Gayatri Balasamy only blurred that line. As a result, parties must remain wary: even “final” arbitral awards may yet be reopened under one of the nebulous exceptions the Court identified.
True reform lies not in creative jurisprudence but in legislative clarity. The Supreme Court itself has observed that meaningful change on appeals and modifications awaits Parliament or the Act’s rules. Until then, Gayatri Balasamy will leave Indian arbitration caught between a stated preference for finality and the reality of ongoing court supervision. In highlighting this gap, commentators have urged viewing the decision not as an end to debate, but as a prompt for legislative fix. Without such reform, judicial overreach will remain the “lingering issue” in Indian arbitration, hardly the outcome parties seeking a swift, autonomous process bargained for.