There is an old saying, “to err is human”. But what about machines? Is humanity imposing double standards on artificial intelligence?
COVID-19 has wreaked havoc of magnitude proportion across the globe, not even machines had been spared of its ramifications.
This is exemplified by an incident back in April 2020 when traders attempted to sell off May’s oil future contracts which resulted in a futures oil prices nosedive into negative digits (Note: these are crude oil we are speaking of, so normal consumers will have little use for such oil – you cannot store it in your backyard either).
On 20 April 2020, the price of West Texas Intermediate futures contracts traded to as low as negative $40.32 per barrel.
Dangers of Machine Reliance
Investors trading on the algorithmic trading platform Interactive Brokers were among those who took a major hit that day, but for a slightly unusual and peculiar reason.
In a nutshell, the brokerage firm’s platform was NOT programmed to trade on negative price. Its users mistakenly placed and were consequently locked up with positive bids while the price went below zero with insufficient equity in their accounts to satisfy the maintenance margin. The technical blip had caused Interactive Brokers to suffer US$88Million as it accepted responsibility for any loss suffered by its users while futures price went negative.
This costly oversight raises an interesting issue on how the law allocates loss occasioned by erroneously executed automated contracts.
But under existing laws, should Interactive Brokers be obliged to accept responsibility?
The Singaporean Decision
In the B2C2 Ltd v Quoine Pte Ltd  SGHC(I) 03,  SGCA(I) 02 the question of machine liability was explored by the Singaporean Courts.
Quoine operates an algorithmic virtual currencies trading platform (“Platform”) programmed to place and close orders without human intervention. B2C2 is an electronic market maker, which trades virtual currencies on exchange platforms with the aid of its algorithmic trading software (“Trading Software”).
In April 2017, B2C2, through its Trading Software, placed orders on the Platform to sell Ethereum for Bitcoins. As a result of several technical errors on the Platform, it erroneously sold Bitcoins held by its two margin traders (“Margin Traders”) to B2C2 at a significant undervalue in favour of B2C2 (“Transactions”). The Platform automatically credited the proceeds of the Transactions to and debited the corresponding Ethereum from B2C2’s account. Upon discovery of the error, Quoine reversed the Transactions.
B2C2 commenced proceedings against Quoine at the Singapore International Commercial Court (“SICC”), claiming reversal of the Transactions amounted to a breach of contract and trust by Quoine, and the matters went to the Singapore Court of Appeal (“SGCA”).
We will focus on the courts’ deliberation on the issues of implied term and unilateral mistake.
No Implied Term Allowing Operator to Unravel for Technical Error
The platform agreement between Quoine and B2C2 contains a clause that once the order is completed on the Platform it is irreversible.
The SICC held, and the SGCA concurred, that no terms can be implied allowing Quoine to reverse any transaction whatever the circumstance. It reasoned that the implication of such a term undermined the very certainty that the agreement sought to instil in algorithmic trading on the Platform.
Knowledge in Unilateral Mistake if Computer Programme is Responsible
Quoine argued the Transactions should be void for unilateral mistake on the part of the Margin Traders under common law and equity. To find a unilateral mistake, Quoine must show that there was a fundamental mistake on the terms of the contract and B2C2 in seeking to enforce the Transactions must have knowledge of the mistake.
In the context of algorithmic trading, SICC held that algorithmic programmes “have no mind of their own” and likened them to “a kitchen blender relieving a cook of the manual act of mixing ingredients”. Further ruling that the intention or knowledge underlying the mode of operation of a robot or trading software would be that of the person who was responsible for causing it to work in the way it did, in other words, the programmer, at the time when the programme was written.
SICC held that B2C2’s programmer who designed the Trading Software did not have actual knowledge (to invoke unilateral mistake under common law), nor constructive knowledge (to invoke unilateral mistake under equity) of such mistaken belief.
The SGCA further found that the Margin Traders were not under a mistaken belief as to the term of the contract, but had mistakenly assumed that the Platform would not fail, which was not an actionable mistake under common law.
The defence of unilateral mistake at common law raised by Quoine therefore fails.
So the question comes again, was Interactive Brokers obliged to accept responsibility under existing laws?
Courts in B2C2 Ltd v Quoine Pte Ltd adopted an incremental approach to develop the law in novel context and confined their legal discussion to deterministic programmes, thus leaving open contracts formed between more sophisticated artificial intelligence which may be said to “have a mind of their own”.
Given that the law is still at its embryonic stage, institutions and platform operators of algorithmic trading are encouraged to make adequate provisions for risk allocation in the event of technical glitch or abnormal trades in their service contracts, and implement robust monitoring to identify and contain any technical glitch in time. Hong Kong Monetary Authority's recent circular titled “Sound risk management practices for algorithmic trading” published on 6 March 2020 may provide some helpful guidance.
This article first appeared in Hong Kong Lawyer.