Lord Chris Holmes of Richmond MBE was a highly successful Paralympian winning nine golds, seven silvers and bronze at Paralympics in Barcelona, Atlanta, Seoul, and Sydney. He was non-executive director at the Equality and Human Rights Commission in the UK where he headed up the disability committee. Chris contradicts the idea that politicians know little about technology. As an active member of the UK’s second house (the House of Lords) he sits on the highly influential cross-bench Science and Technology select committee, and this year his private members’ bill – The Artificial Intelligence [Regulation] Bill has been introduced, with its second reading scheduled to move through the UK Parliament in 2025.
The Bill has been drafted with the principles of trust, transparency, inclusion, and accountability running through it, and would, if enacted in a form close to that of its intended purpose, provide a legal definition of AI. It calls for the creation of an AI Authority tasked with coordinating regulators, includes the accreditation of AI auditors and the promotion of AI literacy, and covers regulatory sandboxes and the appointment of AI responsible officers.
We asked about developing a technology industry in the UK that is inclusive and involves people in all sorts of walks of life, especially in AI.
“When we did our House of Lord Select Committee report on AI, first published in 2018, what rings absolutely true to this day from that report was our headline point of ethical AI. We have a particular opportunity in the UK to develop and deploy ethical AI because of our historical and philosophical point of view, because of our great good fortune in having English Common Law and the principles and structures which underpin it. But that’s the UK. So what we’d need is to have a convening power, not in any sense imposing, but convening, gathering, discussing, and enabling this understanding: That, by having an ethical bedrock to AI development, by having it inclusive by design, it doesn’t matter which polity, it doesn’t matter which jurisdiction you’re in, it would be beneficial to all.”
There has been a certain historical bias in the learning bodies given to AI and LLM models that’s due in part to the data that’s publicly available, and in part down to the demographic of the people who typically build technology. That’s a consideration that Chris acknowledges. “So if you have a large language model, for example, which crawls over the internet to train itself, you’ll get particular answers, particular views, particular perspectives, particular biases by the nature of what is on the internet,” he said.
“The web, crucially what isn’t on the web, for example, as in Sub-Saharan Africa: How much of a view of their perspective is currently accessible and available? So you’ve got two key issues which need to be very carefully considered. Firstly, being incredibly cognisant of the consequences of that data and its limitations. Secondly, are you respecting all of the IP and copyright obligations that you absolutely must? In my view, I don’t see that it’s in anybody’s benefit other than the few players who would do this. If we have a winner-takes-all approach to this, that will suit a very few people. But it can’t be right that there is an argument made that because this is novel and cutting-edge, that all of the conventions, principles, and legal doctrines somehow don’t apply.”
The blockchain is technology that continues to struggle to gain mainstream transaction, although it can be found making gradual appearances in areas like finance and supply chains. Lord Holmes was instrumental in steering the Electronic Trade Documents Act through Parliament until it was successfully enacted in 2023. It covers the establishment of a system of record for digital trade documents based on a blockchain.
“There are two strands worth considering about the Electronic Trade Documents Act [ETDA]. Firstly, all of what it will enable in the trading ecosystem. Secondly, what it says to us in terms of how we can legislate for new opportunities from emerging new technologies, however we choose to define them. [The ETDA] is a blockchain bill that rightly and effectively never mentions blockchain.[…] In many people’s minds their view of blockchain could be paraphrased as blockchain equals Bitcoin, and thus is speculative and unfortunate, and we shouldn’t go anywhere near that. So the Bill is smart in that it doesn’t mention blockchain [and] doesn’t mention any technologies. I’ve described it as the most important law no one has ever heard of, and it is because it does give us that sense of how we can effectively […] legislate for the new opportunities from these technologies. So of itself, it’s a very neat and effective statute, but it demonstrates what we can do and how we now need to have a flow of these types of Bills coming through [Parliament] across a whole series of areas.”
While the Electronic Trade Documents Act is a quiet yet effective way that authorities accept digital documentation (in this case, para-governmental organisations like HMRC and Customs & Excise), the deployment of AI has yet to be legislated about comprehensively by national governments – probably because of the relative speeds of legislative timetables and technological innovation. Lord Holmes thinks that “if we’re not at the top of the AI hype cycle, we are pretty close to it. But we need to always consider all of the technologies that we have in our hands. I believe we should conceive of them as tools. They’re tools in our human hands and thus, we decide, we determine how they’re deployed, for what purpose and for whom. We are in control. We should always know that it’s human-led technology.”
Part of the issue that many people invested in technology have with AI is that the nature of the learning data fed into large models is opaque, and that the open-sourcing of (some) algorithms is thereby largely irrelevant. Chris told us that “transparency, [and] openness are incredibly important because that then leads you neatly into accountability, assurance, and all of those [types of] concepts. I don’t say that necessarily the technologies are neutral of themselves. […] What we need to ensure all from of these technologies is that there is that quality of transparency, of accountability, of assurance, of accessibility, of inclusive by design. […] We have the legal principles, the ethical framework, the philosophical, the social, the economic, the political. We have what we need to make a success of these technologies, and if any of this goes wrong, that essentially won’t be a failure of the technology, it’ll be a failure of us.”
At the time of writing, the Data Protection and Digital Information Bill is in the Committee stage in the House of Lords, undergoing discussion and amendment proposals [PDF] that will eventually form UK law. It contains rules, role definitions and procedures which define the basis on which organisations and individuals interact digitally in the UK and further afield. All laws need to be fantastically detailed, and difficult definitions clarified. For example, “the sense of what does that ‘collective [interest]’ mean, I think, is where we’re going to have some of the biggest and most detailed conversations, particularly when we get to Committee stage of the Bill,” he said.
The wheels of government may appear to turn slowly, in contrast to the flywheels of technological development, but necessary intelligent discussion and research take time. ‘Move fast and break things’ isn’t an ethos that translates from software development to legislative processes. And there are a good number of reasons why that’s a good thing.
