Home Office to deploy AI age-checks it knew were flawed

TL;DR:

  • A leaked Home Office report found its AI facial age-estimation tool wrongly judges teenagers to be adults, with high error rates for Sub-Saharan African migrants.
  • The department tried to withhold the report and disbanded its scientific advisory committee, advisers say, to avoid scrutiny.
  • Facial Age Estimation will still be used at the border from next year, under a £322,000 contract.

The Home Office knew an AI tool intended to check the age of small-boat migrants was flawed but pressed ahead with its rollout anyway, an investigation by The Independent, Lighthouse Reports and WIRED has revealed. A leaked internal report — which the department tried to withhold via a freedom of information request — found the facial age-estimation system judges teenagers to be adults in error.

‘Baked-in’ bias

The report, produced in April 2025 by civil servants running the biometrics programme, found the technology least accurate for migrants from countries such as Eritrea and Sudan, now the top nationalities crossing the Channel. Error rates for female child migrants from Sub-Saharan Africa averaged 4.6 years, meaning a 14-year-old girl could be classed as an adult. A separate audit found the chosen provider’s tool misclassified more than a third of 16-year-olds as adults — rising to nearly 70% on lower-quality border images.

Scientific advisers spoke out for the first time, with Professor Tim Cole of UCL calling the system “hideously inaccurate” and suggesting the Age Estimation Science Advisory Committee was stood down to avoid damning scrutiny. Some 60 organisations have written to the Home Office demanding a halt, warning children should not be “test subjects” for biased technology. The £322,000 contract has gone to Akhter Computers, using an algorithm from German firm Cognitec.

Looking forward

The case is a stark UK test of AI governance in practice: a department deploying a system its own evaluation flagged as discriminatory, having sidelined the experts best placed to challenge it. It contrasts sharply with the reassurance offered elsewhere in Whitehall, where the ICO has set out a plan for “safe” AI innovation. With the rollout due next year, the unresolved question is whether ministers face any binding check on deploying high-stakes AI on vulnerable people — or whether political urgency simply overrides the evidence.

Share this article