TL;DR:

  • Home Office junior minister Lord David Hanson has told Parliament that AI age estimation has been tested against approximately 2.5 million images of people with known ages, across ethnicities, genders and age ranges.
  • Hanson said early assessment suggests facial age estimation is “effective” and could produce workable results faster and more cheaply than bone X-rays or MRI scans.
  • Further tests are planned through 2026, with earlier ministerial suggestions that the technology could be “fully integrated” into immigration processes by year-end.

The scale of the test — 2.5 million images — is a material increase on anything publicly disclosed about UK age-assurance deployment to date. The admission sits inside a thicket of policy work running in parallel: the Immigration White Paper, the Online Safety Act’s age-assurance regime, and the ICO/Ofcom joint statement on age-assurance technologies earlier in the quarter.

Cost and speed, not accuracy

Hanson’s framing in the written reply pivots almost entirely on operational economics. Facial age estimation can be run “at a fraction of the cost” of MRI or bone X-rays, he said, and “much quicker.” That is a different evaluative lens from the Online Safety Act context, where the statistical reliability of age inference — especially for the 16–21 age band — is the debate that matters most for Ofcom’s supervisory approach. UK vendors active in this market, including Yoti and VerifyMyAge, have historically emphasised per-band accuracy rather than per-check cost.

What’s missing from the disclosure

The minister did not disclose which industry algorithms were tested, performance by demographic segment, or false-positive/false-negative rates at each age threshold. That is a significant omission. The Age Appropriate Design Code and ICO guidance on biometric processing both require fairness and accuracy assessments by protected characteristic — data the department has not yet shared publicly.

Looking forward

The earlier ministerial claim that age estimation “could be fully integrated” into immigration processes by end-2026 now has a concrete testing trail behind it, but not yet the published evaluation evidence. Civil society groups and the Lords Justice and Home Affairs Committee have both asked for published test methodology. The next parliamentary update, expected later in the year alongside Further Immigration Reform legislation, will be the key moment to watch — both for the published accuracy data and for whether safeguards against deployment at scale have hardened.

Share this article