Why courts should be given the power to modify awards in domestic arbitration

Section 34 of the Arbitration and Conciliation Act, 1996, which provides for a recourse to parties where they are aggrieved by an arbitral award, needs legislative reform.

By amending the Arbitration Act, courts ought to be given the power to modify - and not just set aside - an arbitral award (rendered in an India-seated domestic arbitration) when it falls foul of public policy. While that is an uncomfortable proposition, I argue that it may be the best alternative to the problems that plague the Section 34 regime.

It has been a little more than a year since the Gayatri Balasamy judgment - on the question of a court’s power to modify an arbitral award - was rendered and it may be a good time to revisit how that judgment has played out in practice. For starters, there is inconsistency in the text of the majority opinion itself. On the one hand, it read the power of modification into Section 34 and also recognised that modification would be a better solution than have parties re-arbitrate again when an award is being set aside. Having said so, in the operative part and conclusions, the power to modify an award was then severely curtailed and limited to categories such as clerical errors, corrections in computational errors, typographical errors, as well “as other manifest errors” provided there is no merits-based evaluation, post award interest.

In practice, the Gayatri Balasamy categories for modifying an arbitral award cause more confusion than resolution. The categories are narrow, ill-defined boxes, incapable of precise definitions. For instance, a computational error can be a very broad ground or a very narrow ground depending on the eye of the beholder and proclivity for computation [see for example Saisudhir Energy v. NTPC Vidyut Vyapar Nigam Limited (2026)]. Ultimately, this results in inconsistent judgments on which cases qualify for modification.

Also, as there is now a very limited power available to the Section 34 court to modify an award, there has been a surge in cases where a prayer is made directly before the Supreme Court to modify an award on grounds other than the Gayatri Balasamy ones, using Article 142 of the Indian Constitution. This is because the Supreme Court alone has the power “to do complete justice”. This has indeed happened in the Bhupesh Bhayana case (2026).

This practice has terrible consequences as a party will ultimately never be certain as to what the ultimate outcome will be (and whether an award will be modified eventually) because the Article 142 power is only with the Supreme Court. Businesses rely on certainty and predictability, and are loathe the use of unbridled discretion, especially where such discretion can only be pressed into service a stage from where no further appeal lies.

Therefore, some form of legislative reform is the need of the hour that preserves the best of the current system, but also addresses the practical difficulties addressed above.

The threshold for interference with awards must not be whittled down. The Section 34 court should not become the first appellate court akin to one that hears appeals from decrees. Instead, keeping the very high threshold for interference in tact, once the court finds that an award must nonetheless be set aside, the court must then be given the power to modify the award.

In other words, the salutary principles of Section 34 as it now stands, should be retained. These include, inter alia, -

a. Limited interference on grounds already specified in Section 34: no review on the merits of the dispute; patent illegality must go to the root of the matter; no-interference merely because there is an erroneous application of the law; no re-appreciation of evidence.

b. The arbitral tribunal and not the court is the ultimate master of the facts and quantity and quality of evidence.

c. If the view taken by the tribunal is a plausible one, the award must be upheld.

d. A matter of construction of terms of the contract is within the domain of the arbitral tribunal alone.

e. Severance: If the non-offending parts of the award can be upheld, they ought to be severed from the offending parts; the whole award need not be set aside.

One may rely upon the judgments in Associate Builders v. DDA (2015) read with Ssangyong Engineering & Construction Co Ltd v. NHAI (2019), which capture the Section 34 remit comprehensively.

But the high threshold of interference by itself is not enough.

The Section 34 scheme does not provide a desirable solution in cases where the high threshold for interference is met and the court concludes that the award must be set aside on grounds of public policy. For example:

(i) When the court reaches a conclusion that an award ought to be set aside as has ignored substantive law [see SEPCO Electric Power Construction Corporation v. GMR Kamalanga Energy Ltd (2025)]; or

(ii) When the court finds that the award must be set aside as it has ignored the terms of the contract, or worse, re-written them. 

As the law presently stands, in such cases, the only recourse available after the award is set aside is to re-arbitrate the dispute all over again [see the McDermott judgment (2006)]. While that is indeed a legal recourse, it is a totally inappropriate and undesirable one, leaving parties in an endless loop. This is because a judgment that sets aside an award will obviously be subject to an appeal. Assuming that it takes several years for that Section 37 appeal to be decided and then a special leave petition (SLP) to the Supreme Court from that judgment be decided, is it to anybody’s benefit that parties re-arbitrate the matter again and then follow the same rigmarole? Would it not be better if the Section 34 court were to modify the award? In our example of where the award is set aside because the award has re-written the contract, it would be a far simpler and more desirable solution for the Section 34 court to apply the terms of the contract instead and modify the portion of the award that rewrote the terms.

There is lack of confidence in the arbitration ecosystem amongst some players, not least the government and PSUs. In recent times, parties have been recalcitrant to resolve disputes by arbitration (see for example the MoRTH Circular dated January 12, 2026; or the Ministry of Finance Guidelines on Arbitration and Mediation clauses on contracts for domestic procurement dated June 3, 2024).

Giving courts the power to modify an award albeit maintaining the very high threshold for interference will only inspire confidence in the naysayers, as ultimately, they will have the comfort of having an award modified if it is so egregious as to violate public policy.

The UK Arbitration Act, 1996 allows courts to vary awards (see Sections 67- 69). That legislation itself is a sui generis legislation and not strictly based on the UNCITRAL Model Law. Even the Singapore Arbitration Act of 2001 allows courts to vary awards (see Section 49) in case they are domestic awards. Surely, the power granted to the courts to vary an award has not affected the reputation of London or Singapore as a seat of arbitration. Why should a provision like that by itself affect India’s reputation as a seat?

Amending the arbitration law is sometimes necessary. The recent amendments to the English Arbitration Act in 2025, that largely undid the impact of the Enka judgement [2020]of the UK Supreme Court, shows that sometimes a legislative amendment may be necessary to preserve what is best for an arbitration ecosystem to thrive.

The legislative solution allowing modification does not do any disservice to the best practices that the Section 34 regime currently has. It will also not add to taking up more court time, as these matters are anyway not decided on leave/by circulation and are heard at length. Moreover, the amendments will not apply to foreign awards as the power to modify will only be restricted to domestic India-seated awards falling foul of public policy under Section 34 of the Arbitration Act.

Sourav Roy is an advocate-on-record at the Supreme Court and Solicitor (England & Wales).