Viscount Younger of Leckie (Con)

My Lords, there is a rather gloomy atmosphere here, but I am not quite sure why. My remarks will be relatively short. I find myself in a very unusual position—namely, I offer strong support for Amendment 124A tabled by the noble Baroness, Lady Bennett of Manor Castle. I do so not only because it incorporates vital safeguards but because it speaks to a principle that these Benches have highlighted and pressed for throughout Committee: that powerful tools must be matched by proper protections. I think we all agree with that. This amendment could not be timelier. The use of artificial intelligence and automated systems is rapidly expanding across Whitehall, with departments increasingly deploying these tools to assist them in undertaking administrative tasks. There are clear benefits to this: efficiency, consistency and the ability to process large volumes of data quickly. AI can be a force multiplier. It can relieve overstretched teams and streamline basic tasks—I saw that when I was in post in the department—but it can never be a substitute for fair and human decision-making where individuals’ rights, entitlements and welfare are concerned. The temptation to lean too heavily on automation is very real, particularly in areas such as social security where volumes are high and budgets are stretched. We have sought to highlight several times to the Government the additional workload and expense that we believe the provisions in this Bill will introduce for the department. Once we incorporate the need to consider additional needs, disabilities and those at risk of coercion—important safeguards that noble Lords across the Committee have supported—we start to face a massive workload. It is feasible, in light of this, that AI will increasingly be incorporated as part of this process, but we must ensure that this temptation is tempered by caution, principle and foresight. This amendment does just that; it makes clear that automation can assist, but not replace, the human judgment at the heart of a fair welfare system. Let there be light. We are not legislating simply for this year, or even this Parliament. We are legislating for a system that must hold up under future Governments, under future pressures and in a future where Al capabilities are likely to expand even further. In just the past couple of years, we have all seen how dramatically these technologies have entered into our lives, often with little warning and even less scrutiny. The safeguards that we write into this Bill now are therefore not merely reactive, they are pre-emptive, and they are essential, a fact that groups such as JUSTICE have recognised and highlighted to us. That is why we have tabled our amendment with the same intent and near-identical wording. It is a proposal that we support wholeheartedly, and I commend the noble Baroness for bringing it forward at this stage. The amendment would require four simple, yet fundamental things: first, that there is meaningful human involvement in any decision-making process that includes an automated element; secondly, that the individual affected receives an individual explanation, including how automation impacted their case; thirdly, that they are given a clear opportunity to make representations; and, fourthly, that they are provided with accessible information on how to challenge the decision. These are not high bars; they are the basic hallmarks of a just and humane administrative process. There are also some important questions around accountability here. If there are no controls in the Bill on how AI is used, there is nothing, it seems to me, that would stop the department introducing this further as a matter of operational efficiency. However, this would have massive implications for the review process, which we have rightly discussed at length during Committee. If a decision is even partially informed by AI, who is held accountable? Could the civil servant in question blame AI instead of taking responsibility? These are serious questions, and without proper safeguards in the Bill, we have no assurance from the Government that we could not, in the very near future, have a situation in which a person is attempting to review a case in which a mistake was made where the fault lies at the feet of a computer program, to put it bluntly. If we have clear human involvement in this process—guaranteed, not just promised—at least there is a person included in determining the final decision who can be held to account. This is a vital safeguard upon which the entire review mechanism would rest. I can anticipate the response from the Minister: she will say that a human will always be at the end of a decision. However, it is not future-proofed, and I urge her to reflect on the long-term value of this amendment and to recognise that it would strengthen the Bill not only for today, but for the years to come. If the Minister can demonstrate to the Committee that these concerns will be protected against not only now, but in perpetuity—which is, of course, the effect of legislation when passed—I would be most grateful. However, from my perspective, I fear the Minister would struggle to meet this challenge because of how the Bill is drafted. I therefore believe there would be real value in the Government adopting this amendment to make sure that they, and the people they serve, are protected not only now, but into the future.