Viscount Camrose (Con)

Many thanks to the noble Lords who have spoken in this debate and to the noble Lord, Lord Freyberg, for his Amendment 60. Before I start, let me endorse and add my name to the request for something of a briefing about the AI Bill. I am concerned that we will put a lot of weight of expectation on that Bill. When it comes, if I understand this right, it will focus on the very largest AI labs and may not necessarily get to all the risks that we are talking about here. Amendment 60 seeks to ensure that the Bill does not allow privately funded or commercial activities to be considered scientific research in order “to avert the possibility that such ventures might benefit from exemptions in copyright law relating to data mining”. This is a sensible, proportionate measure to achieve an important end, but I have some concerns about the underlying assumption, as it strikes me. There is a filtering criterion of whether or not the research is taxpayer funded; that feels like a slightly crude means of predicting the propensity to infringe copyright. I do not know where to take that so I shall leave it there for the moment. Amendment 61 in my name would ensure that data companies cannot justify data scraping for AI training as scientific research. As many of us said in our debate on the previous group, as well as in our debate on this group, the definition of “scientific research” in the Bill is extremely broad. I very much take on board the Minister’s helpful response on that but, I must say, I continue to have some concerns about the breadth of the definition. The development of AI programs, funded privately and as part of a commercial enterprise, could be considered scientific, so I believe that this definition is far too broad, given that Article 8A(3), to be inserted by Clause 71(5), states: “Processing of personal data for a new purpose is to be treated as processing in a manner compatible with the original purpose where … the processing is carried out … for the purposes of scientific research”. By tightening up the definition of “scientific research” to exclude activities that are primarily commercial, it prevents companies from creating a scientific pretence for research that is wholly driven by commercial gain rather than furthering our collective knowledge. I would argue that, if we wish to allow these companies to build and train AI—we must, or others will—we must put in proper safeguards for people’s data. Data subjects should have the right to consent to their data being used in such a manner. Amendment 65A in the name of my noble friend Lord Holmes would also take steps to remedy this concern. I believe that this amendment would work well in tangent with Amendment 61. It makes it absolutely clear that we expect AI developers to obtain consent from data subjects before they use or reuse their data for training purposes. For now, though, I shall not press my amendment.