B2B Artificial Intelligence Transactions: A Framework for Assessing Commercial Liability

24 Pages Posted: 25 Mar 2022 Last revised: 19 Sep 2023

See all articles by Ernest Lim

Ernest Lim

National University of Singapore (NUS) - Faculty of Law

Multiple version iconThere are 2 versions of this paper

Date Written: February 3, 2022

Abstract

Business to business (“B2B”) artificial intelligence (“AI”) transactions raise challenging private law liability issues because of the distinctive nature of AI systems and particularly the new relational dynamics between AI solutions providers and procurers. This article advances a three-stage framework comprising data management, system development and implementation, and external threat management. The purpose is to unpack AI design and development processes involving the relational dynamics of providers and procurers in order to understand the parties’ respective responsibilities. Applying this framework to English commercial law, this article analyses the potential liability of AI solutions providers and procurers under the Supply of Goods and Services Act 1982 (SGSA) and the Sale of Goods Act 1979 (SGA).

The following key arguments are made. First, a stronger case can be made for B2B AI transactions to be characterised as contracts for the supply of services under the SGSA. Further, it is argued that AI software is unlikely to count as ‘goods’ under these legislation. Nevertheless, the article explains why the SGA is still relevant and the analysis will include both statutes.

Second, under the SGSA, whether the provider has supplied the AI system with reasonable care and skill depends on at which stage of the three-stage framework the care and skill has been exercised, and by which party. The framework enables the nature and scope of responsibilities that have been assumed by providers and procurers to be delineated. After all, procurers could be responsible, for example, for selecting or feeding erroneous or insufficient data into the AI system. It is also argued that providers should not be permitted to evade liability by arguing for a lack of causation due to the inherent unpredictability of machine learning.

Third, under the SGA, whether and the extent to which liability will be imposed on providers for the AI system’s unsatisfactory quality will depend on whether providers explicitly and clearly inform procurers that inaccurate, incomplete or non-representative data will make the AI system’s quality unsatisfactory; it will also depend on the clarity and precision by which providers circumscribe their role at the testing and training stage. It is also shown that whether providers comply with the fitness for purpose requirement depends on whether it is reasonable for procurers to rely on providers’ skill and judgement at the different stages of the framework.

Finally, in view of the framework, contractual clauses seeking to exclude providers’ liability under the SGSA and SGA could be construed on the whole as unreasonable under the Unfair Contract Terms Act 1977. The validity of duty-defining clauses is also assessed.

Keywords: AI, Machine Learning, Commercial Law, Sale of Goods Act, Supply of Goods and Services Act

Suggested Citation

Lim, Ernest, B2B Artificial Intelligence Transactions: A Framework for Assessing Commercial Liability (February 3, 2022). Available at SSRN: https://ssrn.com/abstract=4025415 or http://dx.doi.org/10.2139/ssrn.4025415

Ernest Lim (Contact Author)

National University of Singapore (NUS) - Faculty of Law ( email )

469G Bukit Timah Road
Eu Tong Sen Building
Singapore, 259776
Singapore

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
296
Abstract Views
897
Rank
188,835
PlumX Metrics