On 22 November 2023, a Private Members’ bill was introduced into the UK’s House of Lords by Lord Holmes of Richmond (Conservative) titled the Artificial Intelligence (Regulation) Bill (the “Bill”). Private Members Bills are public bills introduced by Members of Parliament and Lords who are not government ministers. The intention of them is to change the law as it applies to the general population, however, they are different from public bills introduced by the government, and generally speaking, the majority of them are not passed and so do not bring about a change in the law directly. However, they are often introduced because of concern about particular issues and to raise awareness, and as a result may sometimes indirectly bring about legislative changes.
i. Content of the Bill
The Bill is interesting in that it seeks to effect legislative change apparently intended to address some concerns arising out of the government’s White Paper on Artificial Intelligence issued earlier this year. This set out the government’s intended approach towards the regulation in the UK of Artificial Intelligence. A prior article of ours1 considered the content of the White Paper and detailed what we saw as certain issues of concern with it.
In very brief terms, the government’s intended regulatory approach as set out in its White Paper - a “light touch” one – was stated as initially not introducing any legislation concerned with the regulation of AI, nor did the government indicate any intention to set up a new AI regulator. Rather, the stated intention was to require existing regulators to ensure that the development and use of AI within their existing spheres of oversight was appropriately regulated having regard to certain key principles.
Establishment of an “AI Authority”
However, the Bill, if passed, would by section 1 require the making of regulations by the appropriate Secretary of State that would provide for the establishment of a body called the “AI Authority” – thereby creating the new AI regulator that was specifically not provided for in the White Paper. Section 1 (2) of the Bill then sets out what would be the functions of the AI Authority and, in short, these would include ensuring that the already existing relevant regulators, whom the government’s White Paper provides shall also in due course be responsible for the regulation of AI, will take account of AI and join up their approach to its regulation as well as undertaking a “gap analysis” to seek to identify where there may be gaps in the existing regulatory oversight in which AI may cause problems.
The AI Authority would also be required to undertake a coordinated review of relevant legislation with a view to assessing its suitability to address the challenges and opportunities presented by AI.
Further, the AI Authority’s function would extend to monitoring and evaluating the overall regulatory framework’s effectiveness and the implementation of principles listed in Section 2 (addressed below) of the Bill. It would assess and monitor risks across the economy arising from AI, and conduct horizon scanning, including by consulting with the AI industry, to inform a coherent response to emerging AI technology trends. The AI Authority would also have to support “test-beds” and “sound box” initiatives to help AI innovators get new technology to market, to accredit independent AI auditors, to provide education and awareness to give clarity to businesses and to empower individuals to express views as part of the iteration of the framework and to promote interoperability with international regulatory frameworks.
As such, the drafting of the Bill seems to have been influenced by a concern that the government’s approach, as set out in its White Paper, which in effect indicated that regulation would simply be conducted by existing regulators already established in certain areas of industry, runs the risk of not being effective or sufficient without some higher authority overseeing the work of the existing regulators. Indeed, this picks up on concerns raised in our prior article where we said that there was a risk that existing regulators may not have time, or indeed the necessary skill-sets/technical knowledge, to regulate AI in their areas alongside the existing regulatory responsibilities which they have. Further, the provision requiring there to be a review by the AI Authority to identify gaps between regulators picks up another concern raised in our prior article, namely that areas of industry may fall between the oversight of different regulators such that the use of AI within that area is in effect unregulated or at least insufficiently regulated.
Regulatory principles
Section 2 of the Bill provides for certain principles, to which the AI Authority must have regard, which are drawn from the White Paper or otherwise commonly recognised as being sensible in terms of the development and use of AI or otherwise not controversial.
Regulatory sandboxes
Section 3 of the Bill requires the AI Authority to collaborate with relevant regulators to construct “regulatory sandboxes” for AI, being arrangements by one or more regulators which allow businesses to test and develop propositions in the market with real consumers whilst ensuring that consumers are protected.
Designated AI officer
Section 4 introduces the interesting and, in the views of this author, extremely sensible proposition that there should be regulations requiring businesses which develop, deploy or use AI to have within them a “designated AI officer”, whose duties would extend to ensuring the safe, ethical, unbiased and non-discriminatory use of AI by the business and to ensure, so far as reasonable, that data used by the business in any AI technology is unbiased (i.e. so as to minimise the risk of bias feeding into decisions of the AI system).
Transparency, IP obligations and labelling
Section 5 is concerned with transparency, IP obligations and labelling, providing for the implementation of regulations by the Secretary of State that would require any person involved in training AI to supply to the AI Authority a record of all third party data and intellectual property used in the training, and to assure the AI Authority that all such data and IP that is used is subject to informed consent as to its use and that there is compliance with all applicable IP and copyright obligations. There would also be a requirement that persons supplying a product or service involving AI must give consumers clear and unambiguous health warnings, labelling and opportunities to give or withhold informed consent in advance and that any business which develops, deploys or uses AI must allow independent or third parties accredited by the AI Authority to audit its processes and systems.
Public engagement
Section 6 is concerned with requiring the AI Authority to implement a programme for meaningful, long-term public engagement about the opportunities and risks presented by AI and to consult the general public and persons it considers appropriate as to the most effective frameworks for public engagement, having regard to international comparators.
Definition of AI
There is thereafter as Section 7 a new and interesting definition of Artificial Intelligence, namely “technology that enables the programming or training of a device or software to (a) perceive environments through the use of data; (b) interpret data using automated processing designed to approximate cognitive abilities; and (c) make recommendations, predictions or decisions; with a view to achieving a specific objective”. It goes on to say that AI includes generative AI, meaning deep or large language models able to generate text and other content based on the data on which they were trained.
ii. Commentary
As stated therefore the Bill appears to be founded on the perception that there were gaps in the government’s White Paper issued earlier this year, and perhaps also some frustration that as yet the proposals in the White Paper have not been advanced as proposed. The White Paper had indicated that by around end of September of this year, the government would have published its response on its White Paper and issued its cross-sectorial principles to regulators, together with initial guidance for their implementation by existing regulators. The White Paper had also indicted that by then the government would also have designed and published an “AI Regulation Roadmap”. As yet, however, the government has not done so. It is hoped, as such, that the Bill, even if not gaining sufficient support to become legislation (although it remains to be seen whether it may do so, and indeed one might well expect there to be a large degree of concern within the legislature as to the government’s light touch approach proposed by its White Paper, perhaps sufficient to lead to some legislative action) may at the very least prompt the government towards swifter implementation of its White Paper proposals.
