The UK's Department for Science, Innovation and Technology (DSIT) has issued initial guidance on how regulators can apply the government's five AI principles within their remits.
DSIT describes the document as the first of three phases of guidance meant to promote "greater coherence in AI outcomes across different regulatory remits where possible." It leaves the choice of tools to each regulator.
The guidance builds on a commitment in the March 2023 AI Regulation White Paper1 to produce support for regulators. According to DSIT, the White Paper set out five principles for regulators to interpret and apply within their remit. The retrieved text names safety, security and robustness, and appropriate transparency and explainability, and also refers to fairness. DSIT says the principles are voluntary and that how they are considered is "ultimately at regulators' discretion." It says the guidance is "not intended to be a prescriptive guide" and is not meant to duplicate, replace or contradict regulators' initiatives or existing statutory definitions of the principles.
DSIT says elements may not apply to regulators that take a "technology agnostic" approach, provided they are satisfied their framework adequately covers issues relating to AI adoption. Where it does not, regulators may update the framework or issue AI-specific guidance.
The document is aimed mainly at regulators whose remits are most immediately affected by AI and who may be just beginning to consider its impacts. It says regulators can establish published policy material on AI setting out the outcomes they expect from regulated firms. Its summary table lists using technical standards among the considerations. On safety, it suggests regulators stress the importance of developers and deployers undertaking risk assessments and mitigations, and consider how they should mitigate cybersecurity risks throughout the AI life cycle.
DSIT says it has started establishing a central function to support regulators' understanding of AI risk, provide expert risk analysis and help identify gaps in existing regulatory powers and remits. It says that in its White Paper consultation response it announced a £10 million package to boost regulators' AI capabilities. The central function will work with regulators "in the coming months" to identify how to use that funding.
Phase one is this document, which DSIT says helps regulators "start considering developing" tools and guidance. Phase two will "iterate and expand" on it, and DSIT plans to issue it by summer 2024. DSIT says phase two will also set out more on what the central function offers. Phase three will identify areas for joint tools and guidance across regulators, including possible multi-regulator guidance.
The guidance text reviewed does not state a publication date, and details from the White Paper consultation response are not reported here.