When does “fraud” make a dispute non-arbitrable? The effect of allegations of fraud on the question of arbitrability was scrutinised by the Supreme Court as early as 1961 in Abdul Kadir Shamshuddin Bubere v. Madhav Prabhakar Oak.

While this question has come to be settled more recently through a series of judgments in Ayyasamy v Paramasivam (2016) and Rashid Raza v. Sadaf Akhtar (2019), read with Avitel Post Studioz Ltd v. HSBC Pi Holding (Mauritius) Ltd (2020), the issue of fraud has been split into two genres – fraud simpliciter and serious fraud. Save for holding that fraud simpliciter does not make a dispute non-arbitrable and that serious fraud could potentially render a dispute unfit for arbitration, case law did not demarcate the specific boundaries of these two variants.

In Ayyasamy, the Court’s analysis hinted that “complexity” and “volume” of evidence could be possible parameters to distinguish the two. While not necessarily adopting the complexity and volume-based yardstick, the Court in Rashid Raza distilled the Ayyasamy analysis to propound a test guided by principles of consent to determine arbitrable fraud from fraud that disturbs the ostensible consent to arbitrate. More recently, the Supreme Court in Managing Director Bihar State Food and Civil Supply Corporation Limited (2025) and Rajia Begum v. Barnali Mukherjea (2026), had the occasion to consider the interplay between Ayyasamy and Rashid Raza. Since Avitel, the Supreme Court has affirmed the law laid down in Rashid Raza as the only applicable test.

This article proceeds in 3 parts – tracing pre-2016 jurisprudence, analysing the post 2016-decisions and evaluating the scope of the exception today – arguing that consent-based tests have, by and large, come to replace the complexity yardstick in recent case-law. 

As early as in 1961 in Abdul Kadir, the Supreme Court considered whether an allegation of fraud would vitiate the arbitration agreement. The Court, after considering previous jurisprudence on this question, attempted to carve out a distinction between fraud and serious fraud, holding,

“…every allegation tending [to] suggest or imply moral dishonesty or moral misconduct in the matter of keeping accounts would not amount to such serious allegation of fraud as would impel a court to refuse to order the arbitration agreement to be filed and refuse to make a reference….”

Abdul Kadir did not lay down a particular standard for distinguishing between allegations amounting to serious fraud from those constituting merely fraud simpliciter. Subsequently, the Court in N Radhakrishnan v. Mastero Engineers. In Swiss Timing v. Organising Committee, Commonwealth Games 2010, the ratio of N Radhakrishnan was found to be per incuriam in view of the binding precedent in Abdul Kadir. The Court also held that the mere existence of a criminal complaint does not automatically render a dispute non-arbitrable.

Ayyasamy v Paramasivam

While Abdul Kadir and Swiss Timing held that only allegations of serious fraud would render a dispute non-arbitrable, there was little guidance as to when allegations were of the nature of serious fraud.  The Court in Ayyasamy undertook a detailed analysis of the contours of arbitrability of fraud. It ruled that “simple” allegations of fraud are not enough to avoid an arbitration clause and that courts may refuse reference only where there are “very serious allegations of fraud” amounting virtually to a criminal offence, requiring consideration of voluminous evidence that “can be decided only by [a] civil court”.

The Court indicated two yardsticks to identify serious fraud – first, the complexity of the fraud (qualitative); and second, the volume of evidence (quantitative). In Ayyasamy, it appears to have been the Court’s view that civil courts are better suited to deal with questions beyond a certain degree of complexity, while also being better equipped to deal with voluminous evidence. The Court, however, was conscious that the mere flagging of fraud related issues cannot be used as a vehicle to avoid binding agreements to arbitrate.

On the first yardstick of complexity, Ayyasamy left room for debate as to whether complexity operates as an independent ground of non-arbitrability or merely as one of the markers of “serious” fraud. At one point, the Court frames the two disjunctively, referring to “very serious allegations of fraud… or where allegations of fraud are so complicated,” suggesting that complexity may alone suffice. Elsewhere, it states that reference should be refused only when allegations are of a “serious and complicated nature,” suggesting a cumulative threshold. The judgment thus leaves open whether complexity is sufficient in itself or only demonstrates the requisite seriousness.

The two-step test in Rashid Raza

The Court in Rashid Raza attempted to lay down a more concrete two-step test to distinguish fraud simpliciter from complex or serious fraud. The Court distilled the holding of Ayyasamy into two working tests:

“First, does the plea permeate the entire contract and above all, the agreement of arbitration rendering it void; or

Second, whether the allegations of fraud touch upon the internal affairs of the parties inter se having no implication in the public domain.”

Applying this two-step test to the facts of the case, the Court held that the instant case would not fall within the purview of serious fraud as allegations pertaining to affairs of the partnership and siphoning of funds would not fall within either of the two tests.

The second prong of the Rashid Raza test - whether the allegations carry implications in the public domain - is best understood as a refinement of Booz Allen, where the Court developed a general classification of non-arbitrable disputes (actions in rem, insolvency, matrimonial, criminal, etc). This in rem rationale was later distilled into the four-fold arbitrability test in Vidya Drolia. Notably, while Rashid Raza partly re-renders the Booz Allen classification, it does not carry forward the complexity-based ouster alluded to in Ayyasamy.

Avitel v. HSBC - an affirmation of the Rashid Raza approach

In 2024, the two-pronged test in Rashid Raza found acceptance in Avitel Post Studioz Ltd. v. HSBC PI Holdings (Mauritius) Ltd, with the Supreme Court handing down a detailed analysis of the development of law on the arbitrability of fraud.

The Court, while reconciling the use of the phrase “serious fraud” in Ayyasamy in light of the test later proposed in Rashid Raza, concluded that “‘serious allegations of fraud’ arises only if either of the two tests laid down [in Rashid Raza] are satisfied, and not otherwise. The first test is satisfied only when it can be said that the arbitration clause or agreement itself cannot be said to exist in a clear case in which the court finds that the party against whom breach is alleged cannot be said to have entered into the agreement relating to arbitration at all.”

The second test requires that the matter involve “questions arising in the public law domain”.

Therefore, Rashid Raza and Avitel limited the test to a qualitative inquiry focused on whether the alleged fraud undermines the arbitration agreement itself or raises public law concerns, thereby drifting away from the complexity-based bar set out in Ayyasamy as a basis for excluding arbitration.

Rajia Begum – Return to the Complexity Test?

Recently, the Supreme Court in Rajia Begum acknowledged the complexity factor laid down in Ayyasamy. In paragraph 12, the Court observes that non-arbitrability may arise “where the court finds that there are serious allegations of fraud which make a case of criminal offence or where the allegations of fraud are so complicated, which need to be decided on the basis of voluminous evidence”.  

However, the Court subsequently adopted the clarification of principles on arbitrability as summarised in Managing Director, which held that serious fraud "is to be understood in the context of facts.” It broke down the Rashid Raza test and clarified,

“The first test is satisfied only when it can be said that the arbitration clause or agreement itself cannot be said to exist…” such that “…the party against whom breach is alleged cannot be said to have entered into the agreement relating to arbitration at all.”

The second test is met in cases where allegations are made against the State or its instrumentalities and the questions raised are not predominantly questions arising from the contract itself or breach thereof, but questions arising in the public law domain.

The Court reiterated that mere allegation of fraud simpliciter may not be a ground to nullify the arbitration agreement, but ruled that when the arbitration agreement itself is alleged to be forged or fabricated, the dispute falls into the “non‑arbitrable” category because it goes to the root of arbitral jurisdiction.

Therefore, Rajia Begum’s reference to the complexity factor is best understood as a descriptive summary of prior case law, while in its own ratio, adopting the refinement to Rashid Raza as articulated in Managing Director.

After Rashid Raza, Avitel, Vidya Drolia, Managing Director and now Rajia Begum, the “serious fraud” exception has been effectively narrowed to two situations: (i) fraud that impeaches the very existence or validity of the arbitration agreement (forged or fabricated clauses, or cases where the party “cannot be said to have entered into the agreement relating to arbitration at all”); and (ii) fraud that raises public law questions, typically involving the State or statutory rights beyond the parties. Everything else - even criminally colourable misrepresentations and siphoning of funds - would be characterised as inter‑partes fraud and is arbitrable.

An isolated departure from the dominant trajectory is seen in how the Delhi High Court ruled in Bentwood Seating System (2025). The arbitrator in his order under Section 16, determined that the adjudication of allegations concerned with fabrication of documents involving foreign companies and public authorities would require assistance better available to civil courts (including through the Ministry of External Affairs), thus treating “complexity” (viewed in the forum non conveniens mould) as reason enough to relegate parties to a civil suit. While the Court attempts to tie the arbitrator’s analysis with the two-prong test by holding that the fraud alleged is of such a nature that it permeates the entire contract and makes out a separate criminal offence, the arbitrator’s own determination appears to rest on the anvil of “complexity”. This approach falls outside the Rashid Raza two-prong framework and illustrates how notions of complexity continue to exert residual influence on arbitrability analysis.

In Ayyasamy, factors of “complexity” and “voluminous evidence” began as pragmatic markers of when a dispute involving fraud might be better suited for trial in court. Through successive case law, the “serious fraud” exception has been doctrinally distilled down to tests of consent (forged or non‑existent arbitration agreements would imply lack of consent) and imperatives in the public domain. Thus, it no longer allows parties to escape arbitration by the mere invocation of complexity, volume of evidence, or criminality. The decision in Bentwood Seating Systems is possibly an infrequent aberration in an otherwise consistent move away from “complexity” as a reason for the ouster of arbitration.

Aditya Chatterjee is a Partner at Keystone Partners.

The author acknowledges the research assistance provided by Kedar Manoj Ammanji, Ameya Krishnaswamy, Ajay Sariel and Mohammad Milhaj.