Abstract:
Non-deterministic artificial intelligence ("AI") systems are black boxes - their programmers lack control and foresight over the AI's outputs. Under the doctrine of unilateral mistakes, the black box problem renders it impracticable to prove that the non-mistaken party ("NMP") who contracts via a non-deterministic AI knew of the relevant mistake. Looking ahead, Parliament should legislate that where the NMP issued the disputed offer through a non-deterministic AI, the contract is voidable if a reasonable person should have known of the mistake at the time of the actual trade. Among other benefits, this rule prevents black box AIs from being abused to circumvent the law on unilateral mistakes and more fairly allocates risk between the NMP and his or her counterparty. The black box problem also means that the unconscionability which grounds equity's jurisdiction in equitable unilateral mistakes will be onerous to prove. Therefore, Singapore's legislature should further codify a rule for this specific situation: that if the NMP acquires actual knowledge of the mistake post-contract formation, and yet seeks to enforce the contract in a manner amounting to sharp practice or impropriety, rescission remains available, subject to the usual equitable bars.