The 126-page Singapore International Commercial Court judgment of Simon Thorley on 14 March in B2C2 Ltd v Quoine Pte Ltd is possibly the first common law judgment about entirely automated algorithmic trading on an OTC dealing platform, and the evident inadequacies of the contractual terms and conditions which governed the relationship between the owner of the platform and its customers.

Quoine owned an algorithmic trading platform where dealers could buy and sell Bitcoin (BTC) and Etherium (ETH) for one another. B2C2 was a dealer on it. On 13 April 2017 Quoine made some software upgrades on the Platform, and it did not do them properly. The result was that the platform stopped working properly and stopped accepting new orders, and consequently the level of trading on the platform became abnormally thin. And Quoine did not notice…at that time, there were two traders, Pulsar and Tomita, who were dealing on the margin. They had borrowed ETH from Quoine and used it to buy BTC 13.52 (worth about USD17,500), which was standing to the credit of their collateral accounts as margin for the loans. To put it another way, they had shorted ETH and gone long BTC, hoping that BTC would go up in value against ETH so that they could then sell the BTC, repay the ETH loan, and make a profit. When the platform stopped working properly, the algorithm determined automatically that Pulsar and Tomita had insufficient margin, and so it issued a margin call and, when they did not comply with it, executed a forced sale routine in relation to the BTC on the collateral account. This meant finding a buyer for it, at the best price available. Enter B2C2. B2C2 had its own bespoke trading software, and it had put into the system offers, which remained open (a) to buy BTC at the rate of BTC10 = ETH1, and (b) to sell BTC at the rate of BTC0.00001 = ETH1 (at a guess, these were to buy or sell up to USD4m-worth of BTC, because the actual order to buy BTC was, as things turned out, for BTC 3,092). These extreme prices (BTC10 = ETH1 was about 1/250 of the prevailing market price) would, in the normal course of events, never be the most attractive, and would never be accepted by a willing and conscious seller. However, in this case, because the platform had stopped accepting new orders, there were very few open orders on the system to buy BTC, and so the sale of the BTC under the forced sale was matched with B2C2’s order to buy BTC 3,092. This was bad for Pulsar and Tomita, but things were about to get even worse for Quoine. What its algorithm should have done was to sell the BTC 13.52 standing to the credit of the collateral accounts to B2C2 for ETH 1.352, and then stop, leaving B2C2’s offer to buy BTC open as regards BTC 3,090.648. However, due to “an architectural flaw” in the platform’s operating system it did not do this. Instead, the system executed a forced sale by Pulsar and Tomita to B2C2 of BTC 3,092. Since Pulsar and Tomita only had BTC 13.52, the system showed them as having a debit balance in their collateral account of BTC 3090.648 – worth about USD4m. Simultaneously, the system credited B2C2 with over BTC 3,092, worth about USD4m, and which it had bought for the equivalent of USD16,000.

All this happened just after 11:30 pm on 19th April. Quoine spotted it the following morning, realised it had to be wrong, and reversed the trades. B2C2 claimed Quoine had no right to do that under the contract between them, and so argued that Quoine was in breach of contract, and that it held the BTC 3,092 on trust for B2C2. B2C2 pointed out that “once an order is fulfilled, you are notified via the Platform and such an action is irreversible”, and Quoine could not point to any clause allowing it to reverse trades done in error, because the T&Cs did not contain any such clause. Quoine’s defences included:

  • the right to reverse the trade was implied;
  • the contract was void for “unilateral mistake” at common law;
  • the contract was voidable for “unilateral mistake” in equity;
  • the contract was void for “mutual mistake” at common law;
  • unjust enrichment.

All this was under Singapore law, but the principles seem pretty much the same as in England. B2C2 was victorious. This seems fair because it had created the faulty automated trading platform and then made it available to the world without any warnings that it might not work properly, and with T&Cs that emphasised its finality, so why should it be able to use the law to wriggle out of the consequences? A clause allowing trades to be reversed within a certain period, and perhaps not making credit balances available until this period had expired, would have made sense. Some invoices are issued with the mysterious suffix “E. & O. E.” and some readers may not know that this stands for “errors and omissions excepted”. It is added to counter an argument that the invoice is final. Even if those had been added to BC2B’s statement of account, Quoine might have had something to hang its hat on. Instead, however, its T&Cs said that “such an action is irreversible”. The implied term was rejected as being inconsistent with the statement in the T&Cs that “such an action is irreversible”. The fairly novel part of this judgment is that the Judge had to consider how one or both parties might have been mistaken or at cross-purposes when the actual contract was generated entirely automatically. The Judge sensibly ducked laying down precise principles (particularly with developments in AI always likely to push the boundaries in the coming years) and concluded:

“The algorithmic programmes in this case are deterministic, they do and only do what they have been programmed to do. They have no mind of their own. The operate when called upon to do so in the pre-ordained manner. They do not know why they are doing something… in my judgment, in circumstances where it is necessary to assess the state of mind of a person in a case where acts of deterministic computer programs are in issue, regard should be had to the state of mind of the programmer of the software of that program at the time the relevant part of the program was written.”

This is yet another example of the global common law consensus, with a Singapore decision being given by an English QC of very high regard in IP law, and essentially applying common legal principles.