UK unveils world leading approach to innovation in first artificial intelligence white paper to turbocharge growth

The Department for Science, Innovation and Technology (DSIT) of the UK government has launched a consultation on a proposed high level regulatory framework for and as regards the use of artificial intelligence (AI). Currently, companies applying AI may be subject to various regulations, e.g. under the Financial Services and Markets Act 2000, consumer protection laws, data protection rules, etc. In this context, the DSIT has provided a nice example outlining the different regulations, a fictive insurance company may have to comply with when applying AI to determine future insurance premiums.
Therefore, the DSIT seeks to create an overarching, high level regulatory framework for the deployment of AI without hindering (financial) innovation and technology and without imposing unnecessary burdens upon (financial) market participants. In fact, the new framework shall facilitate public trust in the (use of) AI, shall drive innovation, shall create economic benefits for users, and shall „strengthen the UK’s position as a global leader in AI“. Upon implementation of the new high level framework, individual regulators such as the Financial Conduct Authority (FCA) would then have to adopt adequate rules to transpose this framework into individual regulatory requirements.
##### The new regulatory framework would consist of the following five key principles or requirements:
(1) Safety, security and robustness: AI systems would have to function in a safe and robust manner to ensure reliability during the entire life cycle of the systems. In this context, individual regulators may have to adopt requirements e.g. as regards the review or testing of the systems or as regards technical standards.
(2) Appropriate transparency and explainability: AI systems need to be „transparent“ and explainable. Firms deploying AI should make this fact publicly available to all stakeholders, particularly though to their clients. Explainability refers to „the extent to which it is possible for relevant parties to access, interpret and understand the decision-making processes of an AI system“. Latter is an important criteria to ensure regulatory compliance of users of a system. To transpose this requirement, regulators may have to adopt certain disclosure and recordkeeping rules.
(3) Fairness: Although the prinicple of „fairness“ is included in many regulations, ranging from data and consumer rights protections laws, to financial market rules, the DSIT specifically states that – in the context of AI – it must be ensured that legal rights of individuals and entities are not undermined AND that the outcomes of the use of AI are neither unfair, nor discriminative. Therefore, regulators may need to provide guidance to firms what this priniple implies in the context of AI and in the context of their specific regulatory domain.
(4) Accountability and governance: Each firm applying AI should be required to develop adequate governance policies and accountability frameworks to ensure adequate oversight over the use and application of AI. Therefore, regulators may have to adopt corresponding legal frameworks to set up corresponding obligations and / or develop guidelines as to their expectations with respect to the establishement of clear and appropriate „lines of ownership and accountability“.
(5) Contestability and redress: Users or persons and entities being subject to a decision based upon AI should always have the right to contest such decision. Additionally, where unfair outcomes or outcomes violating regulatory requirements occurred, such persons and entities should be eligible for redress. As a consequence, regulators such as the FCA may need to adjust their redress schemes and / or their regulatory requirements pertaining to contestability.
Furthermore, the DSIT proposes to implement a new central mechanism to „coordinate, monitor and adapt the framework as a whole“ and to monitor the risks arising from the use of AI. The basic functions and obligations within this mechanism particularly as regards risk governance are depicted in below noted graphic.
Graphic 1: Central (risks) function activities

Finally, the new framework would apply throughout the UK, to all firms operating in UK territory.