Government White Paper sets out UK approach to AI regulation

The question

How is the UK Government looking to regulate AI?

The key takeaway

The UK Government plans to frame the regulation of AI in the UK around five key principles it believes will support innovation and foster public trust in the technology. The Government White Paper AI regulation: a pro-innovation approach, (the White Paper), published in March 2023 was the first step towards developing this new framework. And, in May 2023, we started to see regulator action as the Competition Markets Authority (CMA) launched an initial review of AI models in view of the five principles set out in the White Paper (the May Review) which was quickly followed in June by the CMA’s response to the White Paper consultation (the Response).

The background

AI has been the hot topic since the end of 2022 as phrases like “language models”, “training data” and “machine learning” have become common parlance. However, no AI-specific regulatory mandate exists in the UK. To date, the Government has relied on existing regulators to use their regulatory powers to address AI within their remit. This has created a web of regulation that is meant to cover the use of AI across the whole economy. This approach has inevitably led to gaps in the regulatory framework and uncertainty for businesses, consumers, innovators and even regulators.

The Government has been looking at AI and how it is regulated since it published the AI Sector Deal in 2018 (the Sector Deal). The Sector Deal established government funding for, and set out actions to promote, the development of AI in the UK. A number of papers and reports have since followed, including the AI Roadmap and an independent report setting out recommendations for the Government’s approach to AI.
Following the release of the White Paper, the Government has called on regulators to review the use of AI within their remit to think about how it can promote innovation based on the five key principles.

The White Paper

The Government describes the approach in its White Paper as “flexible”, “pro-innovation”, and “deliberately agile and iterative”, clearly envisaging regulation to develop with AI. There are two key concepts in the White Paper which underpin the overall proposed regulatory framework: the five principles and the new central support function.

The five principles, the Government says, are fundamental to the safe and responsible design and use of AI. They will ensure that:

• AI systems are “safe, secure and robust”
• information concerning the decisions made by AI is “transparent” and “explainable”
• AI systems are “fair”
• those that supply and use AI have sufficient “governance and accountability”, and
• decisions and outcomes produced by AI are “contestable” and “redressable”.

Whilst, for now at least, regulators would not be under a statutory duty to enforce these principles, they would be encouraged to consider them when regulating and setting industry guidance.

The new central support function would tie the regulators together and help close the gaps, according to the White Paper. The central support function’s role would be to monitor the use and effectiveness of the overall framework and incorporate feedback into further iterations of AI regulation. It would also be responsible for supporting businesses and innovators to understand the regulatory regime.

Following an initial period of implementation, the Government plans to review the effectiveness of the framework and take a view as to whether a statutory duty to have “due regard” to the principles needs to be imposed on regulators. For now, at least, the Government believes that not legislating allows regulation to remain agile, flexible and responsive to changes in AI and the market.

The CMA’s review

The aim of the May Review is to develop an “early understanding of the market for foundation models” and to identify the risks and opportunities for consumers and competition associated with the use of AI. The May Review will focus on three themes:

• consumer protection
• competition and barriers to entry within the sector, and
• the impact of AI models on competition in the wider economy.

The CMA has clearly been busy thinking about AI regulation, and the Response, published in early June, was supportive of the Government’s approach to AI regulation, agreeing with the five-principles model and the establishment of the central support function. The Response summarised how the CMA believes the five principles might be applied to its remit and how the principles could support the AI market whilst protecting consumers and competition.

The CMA has explicitly stated that it believes free and competitive markets are fundamental to innovation in emerging markets like AI, indicating that we could expect to see a light touch approach at this early stage. The findings of the May Review will be used by the CMA to identify which principles it feels are best suited to supporting the development of the AI market whilst protecting consumers and competition, ultimately informing how the CMA will implement the approach as set out in the White Paper. The Response has demonstrated clearly that the CMA thinks the White Paper is the right approach, and we can expect to see the findings published following the May Review echoing and building on its Response.

Why is this important?

The White Paper and subsequent regulator guidance will inform the decisions and processes of developers and users of AI technologies in the UK. As demonstrated by the May Review, we are starting to see regulators respond to the White Paper by thinking about what AI means for their sector and how they can implement the five principles. The positive response issued by the CMA may pave the way for other regulators to voice their opinions.

Any practical tips?

The White Paper represents the Government’s first, but very cautious, step towards the regulation of AI. That being said, this is a guiding paper for regulators on how the Government expects them to act with regards to AI and what they should be considering when working within their remit. Equally, businesses that use AI technology now have a clearer understanding of the fundamentals of AI regulation and can use this as a toolkit when developing and using AI.

We can expect to see further developments in AI regulation over the course of this year. Regulators are likely to start to issuing guidance for businesses within the next 12 to 18 months, with the CMA in particular planning to publish a “short report” of its findings early September 2023.

Summer 2023