The Law of Artificial Intelligence
By Matt Hervey & Dr Matthew Lavy
The Law of Artificial Intelligence is a 14-chapter tome about what the law is, and about the application of the law to real-world AI as it exists today. In their Introduction in Chapter 1, the General Editors say that the aim is to give practitioners a thorough grounding in AI, the arising ethical and regulatory problems, and to assist in applying the law to real-world AI tools and applications.
Chapter 2 talks about AI, the technology. It introduces readers mainly to machine learning (supervised and unsupervised) and reinforcement learning, and then mentions three legal difficulties: reasonableness, explainability, and foreseeability. While the portions on supervised and unsupervised learning are not that long, Googling would be needed to better understand the concepts discussed. They go into some mathematics as well, reminding me of a joke during a hearing on damages years ago that lawyers studied law because they cannot do maths.
Hundreds if not thousands of articles and papers have been published on AI ethics and ethical principles. Chapter 3 succinctly captures and clearly explaining the key issues in this burgeoning area. The Chapter starts off by discussing what AI ethics is, what the difference is between ethics and law, and why ethics in AI matters. It then goes into the application of ethics in different contexts, the ethical principles of fairness, accountability, and transparency, and questions to aid the evaluation of AI from an ethical perspective.
Chapter 4 sets out regulatory initiatives in the EU, UK, Germany, France, US, Japan, China, and Singapore. There are helpful observations on regulatory trends relating to concepts like transparency, explainability, and bias, as well as on emerging best practices such as AI ethics policies and committees, impact assessments, and audits.
Litigators would be particularly interested in Chapters 5 to 7, which cover liability for physical damage and economic harm, and professional liability. At this point in time, there are still probably no specific and sufficiently detailed rules governing liability for AI, and little caselaw on the same. However as pointed out by the General Editors at the start of the book, just because there is not much law on AI does not mean that AI exists outside the law, or that well-established legal doctrines do not apply where AI is involved. The common law can adapt to address novel issues of liability as they arise, particularly in the field of tort law.
Chapter 8 takes on the gargantuan task of tackling Intellectual Property issues – patents, copyright, data and database rights, design rights, trademarks, trade secrets, confidential information, and contractual measures.
Chapter 9 discusses data protection and privacy issues in the context of the GDPR and regulatory guidance from the ICO. It discusses the interaction between certain GDPR concepts and AI applications, such as anonymisation and profiling, as well as issues relating to the principles under Article 5. There are sections on other topics such as controller/processor relationships in AI systems, rights of individuals, remedies, and enforcement. As the GDPR and the ICO’s guidance are heavily referenced, this chapter would hit the spot for EU and UK cases, and would be less relevant to practitioners operating in other jurisdictions.
Chapter 10 covers competition law issues in the context of the UK competition regime. Chapters 11 and 12 are short chapters, on criminal law (mainly relating to autonomous vehicles) and smart contracts respectively.
Chapter 13 examines and presents some rather thoughtful views on the use of AI in the criminal justice system. It talks about predictive crime mapping, individual risk assessment, and facial recognition, as well as concerns surrounding the use of these tools. The book highlights that the criminal justice system relies on tools developed by private companies whose driving force is not the rule of law but profitability, and that the concept of the rule of law is alien to the development process. There are also the issues of biases in data used, lack of skills and understanding to deploy tools property, and the false legitimacy awarded to decisions by virtue of them having a basis in statistics. Nonetheless it does not mean that these tools should be tossed out. They can be used depending on whether the significant risks and issues associated with their use, are mitigated.
Chapter 14 examines the use of AI in the legal profession, for example in legal analytics tasks, eDiscovery, contract assessment, and case modelling. This may be more prevalent in the UK, but I think stuck in nascency in Singapore.
Even though this book was published just over a year ago in December 2020, it is the only book I am aware of that discusses AI law, ethics, and applications, in such a comprehensive manner. While the target audience is probably practitioners in the EU and UK, many chapters have general application, or can be analogised to other jurisdictions’ laws.
Information about the book may be found here.
The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of the author’s employer.