Unpicking a Fin(e)tech Mess: Can Old Doctrines Cope in the 21st Century?

The decision of the Singapore International Commercial Court in B2C2 Ltd v Quoine Pte Ltd [2019] will be of interest to Commonwealth lawyers as it grapples with novel questions such as the nature of cryptocurrencies and the consequences of mistakes in automated algorithmic contracting. The case concerned seven trades of two cryptocurrencies, Bitcoin (BTC) and Ether (ETH), on 19 April 2017 between three customers of Quoine Pte Ltd (Quoine) the defendant cryptocurrency exchange operator – B2C2 Ltd (B2C2), the plaintiff, and two margin traders, Pulsar Trading Capital (Pulsar) and Mr Tomita (Tomita), as counterparties. In each instance, B2C2 had sold ETH and bought BTC. On 13 April 2017, Quoine implemented changes to some login passwords on its trading platform for security reasons but failed to implement certain necessary changes to its Quoter Program (QP) that was responsible for retrieving market prices on its platform. This resulted in an ‘abnormally thin’ trading volume and triggered margin calls on Pulsar/Tomita resulting in the placement of market orders to buy ETH at the ‘best available market price’. As QP was malfunctioning, correct market price information was unavailable, and it commenced purchasing the relevant currencies at the lowest available price. At some point, B2C2’s own algorithmic trading software, Pure Quote (PQ), placed orders to sell ETH at 10 and 9.99 BTC, which were filled by QP to complete the margin calls. It was not disputed by either party that the trades overvalued ETH by a measure of approximately 250 times its then market value. Following the trades, approximately 3092.5 BTC were credited by Quoine to and 309.25 ETH were debited from B2C2’s account. When Quoine reversed these trades, B2C2 sued it for breach of contract and trust. Numerous issues were raised but in this post we focus on only three.

First, the court rejected B2C2’s submission that Quoine sat in between B2C2 and Pulsar/Tomita in the trades. However, the court then incorrectly concluded that ‘the answer to this question has no bearing on whether or not Quoine is liable to B2C2 for breach of the Platform contract’ (at [129]). If Quoine is not itself a counterparty to the trades, then it must be acting as agent for the traders for the purposes of matching trades. If so, this colours the construction of the irreversibility clause within the Platform contract, which provided that ‘once an order is filled, you are notified via the Platform and such an action is irreversible’ (at [23]). It cannot be construed to mean that the trades themselves are irreversible but rather the instructions to Quoine as agent are irreversible. This may not be the most natural construction of the clause but it is trite law that a literal construction must yield to commonsense: The Antaios.[1]

Secondly, Quoine argued that in placing sell orders at 250 times the prevailing market value for ETH on behalf of Pulsar/Tomita, B2C2 must have known that such orders could only have been executed by mistake. As a general rule, a party to a contract is bound even though he may have made a mistake in entering into the contract. This is a manifestation of the objectivity principle, which promotes commercial certainty. An exception to this rule is when one party knows of the subjective intent of the other. The challenge was one of applying this familiar doctrine in the unfamiliar circumstances of algorithmically concluded contracts.

In a typical case, the law is clear: knowledge of the other party’s mistake must exist at the time the relevant contract is formed. However, a party relying on automation is typically unaware that a contract has been concluded, introducing a significant wrinkle to the problem. The absence of awareness on the part of an offeror that its offer has been accepted is no obstacle to the contract’s conclusion. This is typically the case for unilateral contracts but is also implicit in Lord Denning MR’s analysis of automated bilateral contracting in Thornton v Shoe Lane Parking [2]. Being unaware of the conclusion of a contract, a contracting party utilising automated contracting is ipso facto incapable of having any actual knowledge of any mistake on the part of its counterparty at the time of contracting, however egregious the mistake. But this would have the effect of immunising any contracting party employing algorithmic contracting from the doctrine of unilateral mistake entirely, which cannot be correct. The court accordingly struggled to identify both a person to whom knowledge could be attributed and a time at which it was appropriate to do so. Quoine proposed two possibilities. First, it proposed that attribution of knowledge should proceed on the basis of a hypothetical meeting of minds. Secondly, it proposed that the algorithms should be treated as the legal agents of such party and attribute knowledge on the basis of the knowledge of the software programmer at the time the software was written. The court rejected the former as ‘wholly artificial’ (at [204]) and proceeded to consider the latter. The court rejected personifying the software in the case as PQ was purely ‘deterministic’ – ‘they do and only do what they have been programmed to do’ (at [208]). Nevertheless, despite rejecting the algorithm’s agency, it applied the thrust of Quoine’s alternative submission that the knowledge and intention of the programmer at the time it was written was relevant (at [210]). This led the court to extensively debate whether B2C2’s trading system was designed to take advantage of the errors of others or to exploit distortions in the market. PQ was designed to run in all market conditions, containing no ‘circuit-breakers’ and would continue operating even in illiquid or distorted markets. Instead, it contained ‘deep prices’ on both the bid (buy) and the ask (sell) side, designed to ensure there would always be prices which the software could draw upon.

This approach is unsatisfactory for two reasons. First, by conflating knowledge with intention or motive, the doctrine of unilateral mistake is transmuted into an attenuated version of deceit. Secondly, it is unrealistic to attempt to attribute knowledge of a future mistake to a past programmer a la Nostradamus. The entire justification for including ‘deep stops’ at the time of programming that significantly diverged from prevailing market prices stems from the impossibility of predicting the future. Given that programmers are well advised to plan for the unforeseeable, it is difficult to imagine how the requirement of knowledge can ever be satisfied on this test. Although the court theoretically retained the doctrine of unilateral mistake for automated contracting, the effect of reasoning in this way is to restrict the doctrine’s practical operation to a vanishing point. A reversal of roles in the classic Singapore case of Chwee Kin Keong v Digilandmall.com Pte Ltd [3] will reveal the rule’s obsequiosity to automation. In that case, the defendant mistakenly advertised a particular laser printer for sale at S$66 (the correct price was S$3,854) over the Internet. Prior to the correction of the error, the plaintiffs ordered hundreds of the same. As the mistake in Chwee Kin Keong was on the part of the party employing automation, there was no difficulty in attributing knowledge of its mistake to its human counterparty. But if the roles were to be reversed, an application of B2C2 Ltd v Quoine Pte Ltd’s time of programming rule would render it practically impossible for a human party to attribute knowledge to its automation wielding counterpart, which is grossly unfair. The correct test must instead be the one the court so quickly dismissed. A contracting party is entitled to waive its right to know of the conclusion of a contract through automation, but such waiver cannot also serve to immunise itself from the effects of any vitiating factor such as mistake.

Finally, the claim in breach of trust is the most troubling even though the court must be correct in concluding that cryptocurrencies can be held on trust.[4] In relying on Lord Wilberforce’s classic definition of a property right in National Provincial Bank Ltd v Ainsworth [5], famously criticised by Gray and Gray as hopelessly circular [6], the court appears to fail to appreciate that personal contractual rights can also be held on trust: Lord Strathcona Steamship Co Ltd v Dominion Coal Co Ltd [7]. The court also elides the quite separate questions of whether a right may be the subject-matter of a trust with that of whether the test of certainty of subject matter is satisfied (at [142]). After all, there is no question that wine, shares, or gold bullion can be held on trust yet the question of certainty of subject matter raged in Re London Wine Co (Shippers) Ltd [8], Hunter v Moss [9], and Re Goldcorp Exchange Ltd (in receivership) [10]. There can be no certainty of subject matter on the facts since Quoine had intended to transfer BTC owned by Pulsar/Tomita to B2C2 but ‘the available BTC balances in Pulsar’s and Tomita’s accounts were insufficient to meet B2C2’s orders’ (at [75]). Even the question of certainty of subject matter in respect of the BTC held by Quoine for Pulsar/Tomita is also clouded by the issue of priority between competing claims. Presumably, the BTC held by Quoine on behalf of Pulsar/Tomita would be held under the same terms as the Platform contract with B2C2, which means that Pulsar/Tomita would have pre-existing equitable interests in these BTC. Given the software malfunction, it seems doubtful that Quoine would be, vis-a-vis Pulsar/Tomita, authorised to transfer their equitable interests to B2C2. If so, any declaration of trust by Quoine in favour of B2C2 over assets it already holds on trust for Pulsar/Tomita must be subject to the equitable maxim qui prior est tempore, potior est jure (the first in time) so that B2C2’s equitable title to the existing BTC must rank behind that of Pulsar/Tomita.