Unjust enrichment is a plausible cause of action for individuals whose data has been collected and used without their consent, to train, develop, or improve AI systems, or which has been sold for such purposes. Disgorgement of profits may be possible in some situations where the defendant has unlawfully collected or used personal data. Gain-based remedies have a number of advantages in this context, including the fact that it may be relatively easy to ascertain the gain, but demonstrating the loss will be considerably harder. However, contractual pre-emption may limit the utility of claims for unjust enrichment.

Financial supervisors have begun to use AI to prevent financial distress, detect fraud and, more generally, for investor protection purposes. Similarly, private parties increasingly rely on AI to decide small claims and arbitration cases. In view of this evolution, this chapter deals with the current use of AI in the financial sector, regulation of and by AI, and, most importantly, AI-driven financial supervision.

This Handbook brings together a global team of private law experts and computer scientists to examine the interface between private law and AI, which includes issues such as whether existing private law can address the challenges of AI and whether and how private law needs to be reformed to reduce the risks of AI while retaining its benefits.

Legal technologies using AI-augmented algorithms to translate the purpose of a law into a specific legal directive can be used to produce self-driving contracts, that is, a contract which instead of relying on a human referee to fill gaps, update, or reform the provisions of the contract, uses data-driven predictive algorithms to do so instead. Self-driving contracts are not simply science fiction; not only are self-driving contracts possible, they are in fact already with us.

The law should be ‘computable’, in order to make retrieval and analysis easier. Computable law takes aspects of law, which are implicit in legal texts, and aims to model them as data, rules, or forms which are amenable to computer processes. Laws should be labelled with computable structural data to permit advanced computational processing and legal analysis.

The legal treatment of autonomous algorithmic collusion in light of its technical feasibility and various theoretical considerations is an important issue because autonomous algorithmic collusion raises difficult questions concerning the attribution of conduct by algorithms to firms and reopens the longstanding debate about the legality of tacit collusion. Algorithmic collusion, namely, direct communication between algorithms, which amounts to express collusion, is illegal. Intelligent and independent adaptation to competitors’ conduct by algorithms with no direct communication between them, which is tacit collusion, is generally legal. There should be ex ante regulation to reduce algorithmic collusion.

Consumers are at the forefront of market uses of AI. There are also myriad consumer uses of AI products. Consumer protection law justifies greater responses where the interactions involve significant risks and relevant consumer vulnerability; both such elements are present in the current and predicted AI uses concerning consumers. Whilst consumer protection law is likely to be able to be sufficiently flexible to adapt to AI, there is a need to recalibrate consumer protection law to AI.

The EU definitions of AI moved from a narrow one to a broad one because of the EU policy which is to govern the phenomenon of AI in the broadest way that includes a wide range of situations. The key contents of the main EU AI documents including the European Parliament Resolution with recommendations to the Commission on Civil Law Rules on Robotics, the Ethics Guidelines for Trustworthy AI, the proposed AI Act, and the recent Proposal for an AI Liability Directive, are examined.

The protection of AI-assisted and AI-generated works causes problems for existing intellectual property law. However, it is doubtful whether the purposes of patent law would be served by granting patents for AI-generated inventions. Further, AI systems are unable to make the creative choices to bring their outputs into the realm of copyright protection. However, with AI-assisted outputs, there may still be sufficient creative choices by the programmer or user to bring the output into the domain of IP protection. AI fundamentally challenges the anthropocentric copyright regime. AI technologies will require us to rethink fundamental concepts within IP law, including, for instance, the standard of obviousness applied within patent law.

The UK Parliament has already pre-emptively legislated for a compensation solution for autonomous vehicle accidents through the Automated and Electric Vehicles Act 2018. The Act is a response to the fact that the ordinary approach to motor vehicle accidents cannot apply in an AV context since there is no human driver. Tort law has previously been subjected to major shifts in response to motor vehicles, and we are again on the cusp of another motor-vehicle-inspired revolution in tort law. However, in legislating for AV accidents in the UK, there was inadequate consideration of alternative approaches.

On AI-assisted adjudication, concerns including biases (such as automation bias, anchoring bias, contrarian bias, and herd bias) and ethical worries (such as human adjudicators ceasing to be decision-makers, excessive standardisation of decisions, and the fact that judges may be pressured to conform to the AI’s predictions) can be addressed. Adjudicators may use AI to assist them in their decisions in three aspects: training and implementation; actual use; and monitoring. Because AI will not be able to provide the legal justifications underlying its predictions, the human adjudicator will have to explain why the AI-generated prediction is legally justified. AI will not replace adjudicators.