The point of this is unambiguously to use technology as an accountability sink. You can't have a human eyeball age and call that a "process." You want a machine to point to instead. Its accuracy or lack thereof is immaterial to anyone who would seriously suggest using it.

As ever, this is the real risk of "AI"; not the technology itself so much as the technology-as-social-construct. A machine oracle we can abdicate decisions to with a facade of neutrality.

In this case, the facade is painting over the underlying motivation which is to reject asylum claims. You could imagine a world in which it is instead used to scan and fast-track claims through an automated and unaccountable process, but the form of the deployment has baked-in the outcome and interests of the powerful. Don't be surprised if there's another automated AI system that totally-pinky-promise-for-sure validates that rich tourists aren't terrorists so they can walk through security unmolested and another system that uses AI to flag "suspicious behavior" for the proles. The outcome is baked in, the AI just provides plausibility and legitimacy.