Border Security, Asylum and Immigration Bill

Debate between Baroness Neuberger and Baroness Butler-Sloss

Report stage
Monday 3rd November 2025 Lords Chamber

[Baroness Butler-Sloss (CB)]

I understand very well what the noble Lord, Lord Harper, is saying, but one of the problems, it seems to me, is the differing maturity of children in different parts of the world.

Several years ago, I went to the charity Safe Passage, which has a drop-in centre in north London. I met two Afghan boys who were both truly identified as 16; Safe Passage was absolutely satisfied they were 16, and they actually had some papers to prove it. One of them had a beard and the other had a moustache. Anybody who did not know about different maturity in different parts of the world would take it for granted that they were over 18. There is an added problem here: we need to recognise the differing maturity of children from different parts of the world.

[Baroness Neuberger (CB)]

My Lords, I support Amendment 57, to which I have added my name. I too thank the Refugee and Migrant Children’s Consortium for all the help that it has given us. I also support Amendment 27.

For very good reason, and not for the first time, Amendment 57 would introduce statutory safeguards for individuals whose age is disputed. To the noble Lord, Lord Harper, I say: we do not suggest that we should prohibit visual assessments at the border. What Amendment 57 would ensure is that those assessments comply with child protection principles, especially the benefit of the doubt standard established in case law and international guidance. This principle requires that where age is uncertain, the individual should be treated as a child unless there is compelling evidence to the contrary. That is the principle which I believe we should stick to.

The amendments align with recommendations by the Independent Chief Inspector of Borders and Immigration, as the noble Baroness, Lady Lister, has already said. Crucially, the amendment also addresses the Government’s proposal to use AI-based facial age estimation. I feel that I am a broken record on the subject of facial age estimation, and indeed on age estimation in general. We have had to contend with the proposal to use X-ray systems to determine age, and time after time we have argued that not only is it inaccurate—a point made clearly by the noble and learned Baroness, Lady Butler-Sloss—to use teeth or wrists for X-rays but it is unethical to expose people to unnecessary radiation and that X-rays should be used only for the benefit of the people concerned. We are delighted that the present Government are not proposing X-rays among their scientific methods, and we are also immensely grateful to the Minister for having conversations with us on this subject.

However, the AI systems suggested are not foolproof either. Indeed, independent evaluations show that these systems have error margins of between two and four years, as the noble Lord, Lord Harper, said, and they exhibit demographic bias, which is exactly what the noble and learned Baroness, Lady Butler-Sloss, has said—particularly, it turns out, for younger ages and minority ethnic groups. Academic research confirms that children’s faces are harder to assess accurately and that claims of near-perfect accuracy remain unverified. Overreliance on such technology risks replicating systematic errors rather than fixing them, so we will be replacing human error with machine error.

We all recognise that age assessment is complex and cannot be solved by one measure, but we believe that the Government need to listen to experts and adopt safeguards that make the system safer for children. Amendment 57 offers a practical, rights-based solution. It would preserve operational flexibility at the border, reinforce compliance with children’s legislation, and ensure transparency and accountability in the use of technology. I hope the Minister can give us some more details about how the trial of this AI technology will work, and indeed that he can reassure us that it will not be relied on unless it is truly accurate—but it looks as if we are a long way from that.