News: 1757417129

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

UK Home Office dangles £1.3M prize for algorithm that guesses your age

(2025/09/09)


The UK's Home Office is offering £1.3 million ($1.7 million) to developers of age-determining software - a tech it wants to deploy widely across its systems.

A contract tender posted days ago states the government is looking for an "algorithm that can accurately predict the age of a subject."

"Home Office are seeking to procure an algorithm that can accurately predict the age of a subject," the [1]contract's description reads. "This will have multiple use cases for Home Office, an example could/would be to assist in determining the age of those who are encountered without verifiable identity documentation."

[2]

While the tender does not cite any specific examples, the Refugee Council has previously highlighted the dangers of inaccurately assessing the age of asylum seekers, refugees, and migrant children.

[3]

[4]

The charity rebuked the Home Office in a [5]report [PDF] earlier this year for wrongly classifying hundreds of children as adults at the UK border in just the six months leading up to the 2024 general election, which resulted in "a safeguarding failure on an alarming scale."

If children are wrongly judged to be adults, it can lead them to be placed in unsupervised adult accommodation or detention centers, potentially exposing them to abuse.

[6]

The Refugee Council more broadly scrutinized the Home Office's approach to assessing children's ages, saying its proposals to use scientific measures to more accurately determine undocumented children's ages would be "ineffective, expensive, and harmful."

It claimed these new methods would also fail to meet the core problem at hand – flawed at-port assessments.

Furthermore, it claimed the National Age Assessment Board (NAAB), a decision-making function run by the [7]Home Office ostensibly to assist local authorities in determining children's ages, also has a "profound negative impact" on the mental health of those being assessed, wastes time, and costs an unnecessary £1.7 million ($2.3 million) per year.

[8]

The Refugee Council's report was echoed by the Independent Chief Inspector of Borders and Immigration, David Bolt, in a [9]report [PDF] published in July.

[10]Legacy tech blunts UK top cops' fight against serious crime, inspectors find

[11]UK Home Office hikes tech consultant spend to £350M despite pledge to cut costs

[12]Minority Report: Now with more spreadsheets and guesswork

[13]UK expands police facial recognition rollout with 10 new vans heading to a town near you

In it, Bolt noted that the current methods of determining children's ages at the UK border "cannot be relied upon."

Bolt's report was published the same day that the Labour government announced a trial of AI technology to increase reliable border age assessments.

Angela Eagle, immigration minister at the time of the announcement, [14]said the government was trialing AI tech, trained on millions of images where an individual's age is verifiable, that produces an age estimate to within "a known degree of accuracy," without specifying exactly how accurate that was.

The primary use of the Home Office's new algorithm will be for immigration purposes, but the winning submission will be one that can be baked into the Home Office's hardware and systems already in place, it said.

It listed age determination as the only use case, but the Home Office oversees matters such as crime, policing, drugs policy, and [15]counter-terrorism , as well as immigration.

The Register asked the Home Office to expand on its ambitions for the technology, but it did not respond.

According to the tender, work shall begin on the algorithm on January 19, 2026, with completion expected on the same day, three years later. ®

Get our [16]Tech Resources



[1] https://www-media.refugeecouncil.org.uk/media/documents/RMCC-Lost-Childhoods-March-2025-1.pdf

[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aMBPFYFhmIvctkmhztbcIwAAAJc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aMBPFYFhmIvctkmhztbcIwAAAJc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aMBPFYFhmIvctkmhztbcIwAAAJc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[5] https://www-media.refugeecouncil.org.uk/media/documents/RMCC-Lost-Childhoods-March-2025-1.pdf

[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aMBPFYFhmIvctkmhztbcIwAAAJc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[7] https://www.theregister.com/2025/04/04/home_office_asylum_system/

[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aMBPFYFhmIvctkmhztbcIwAAAJc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[9] https://assets.publishing.service.gov.uk/media/687f6f9dfdc190fb6b84685e/An_inspection_of_the_Home_Office_s_use_of_age_assessments__July_2024___February_2025_.pdf

[10] https://www.theregister.com/2025/09/09/nca_legacy_tech/

[11] https://www.theregister.com/2025/09/03/home_office_consultant_bill/

[12] https://www.theregister.com/2025/08/16/uk_to_use_ai_to/

[13] https://www.theregister.com/2025/08/13/uk_expands_police_facial_recognition/

[14] https://questions-statements.parliament.uk/written-statements/detail/2025-07-22/hcws885

[15] https://www.theregister.com/2025/08/19/uk_apple_backdoor_uturn/

[16] https://whitepapers.theregister.com/



worth a shot

zxm

someone should submit the leisure suit larry age test. will be better than some of the other enteries they get :-)

Predict???

Eclectic Man

"algorithm that can accurately predict the age of a subject."

Predict: say or estimate that (a specified thing) will happen in the future or will be a consequence of something.

I know that words change their meanings over time, Burt the article headline is more accurate. They need something to assess a person's age accurately, not predict it at some future time.

Re: Predict???

Joe W

Well.... yes and no. From a stats language point of view this works. You make a prediction which then can (in a test) be confirmed or not. It is a prediction of an unknown variable based on the inputs. Nothing temporal is implied.

Re: Predict???

Handlebars

The home office has little to no prospect of getting a definitive value from anywhere else, so in this case the tender is really for an assessment tool.

Re: Predict???

Flocke Kroes

The software can be tested against random images from the training data set to come up with a high estimate of its accuracy.

Re: Predict???

Shirley Knot

With edge cases being woefully erroneous.

Re: Predict???

SnailFerrous

I can predict that in one year's time, they will be one year older*, irrespective of their current undetermined age. That'll be £1.3M, ta very much; used tens and twenties in this wheelbarrow please.

Unless they are dead.

Doctor Syntax

Accurate but with what precision?

Just answering 50 ± 50 for everyone isn't likely to make many mistakes.

"a known degree of accuracy,"

Shirley Knot

I guarantee this person is 61yo +/- 61 years. [At time of post.]

MonkeyJuice

Why are they asking for this now? Are they suggesting that (non GDPR compliant) age verification tech they were sure the free market would magically provide doesn't work?

And are they sure they want to spend all that cash on something that can [1]trivially be circumvented as soon as it is released?

It is almost as if our legislators don't know what they are doing. Anyway, I'm off to write some secure but totes backdoorable encryption. It will use AI to decide if you are a Bad Guy or not, because that's how AI works.

Mine's the one thinking of the children while a stranger photographs your children.

[1] https://www.theregister.com/2017/11/06/mit_fooling_ai/

Handlebars

This seems to be mostly to assess migrants on the south coast. But I could easily imagine one of the OSA ones winning the bid, especially if their owners are friends with MPs.

MonkeyJuice

Wait, so they're saying 'actually works' is not a requirement for the OSA?

I guess I'm just not smart enough to be a politician, because I hadn't realised the two identical requirements of 'age verification technology that actually works' could be so vastly different.

phuzz

There's no way it can work 100% of the time, because there's no way to precisely judge someone's age just based on their appearance.

Presumably the contract will specify the level of accuracy and precision expected. Although if they did that, and then haded the contract to one of their pals touting some kind of AI magic box, then a later government could hold them liable for not meeting targets.So actually, they'll probably say the contracted level of accuracy is a trade secret and that the public aren't allowed to know it.

Very difficult for humans

Caver_Dave

I've stood on the door of Young Farmers events for decades.

I know that it is very difficult to estimate someone's age with any degree of certainty.

With Young Farmers, all members have to show their membership card with photograph that clearly states their date of birth.

And I can say, with certainty, that I (and all others on the door over the decades) can be up to +/- 5 years out with our age estimates, especially so at the younger end of the spectrum (we allow 16+ at our evening events in my county, with pre-designated and signed in, non-drinking chaperones for the under 18s).

Drink icon as it is Young Farmers!

Re: Very difficult for humans

Richard 12

It's straight up impossible!

There's a reason for "challenge 25". The age range 10-25 or so is basically indistinguishable by sight. Ask any teacher.

Any system is going to have massive biases baked in, especially for darker skin. This is well known.

Re: Very difficult for humans

GoneFission

This doesn't really seem to be about building a system that can accurately age-profile people, but rather deferring responsibility and liability for an incorrect determination to a third party.

Re: Very difficult for humans

Brewster's Angle Grinder

Yeah, they're asking to solve what is an unsolvable problem. AFAIK there is no reliable connection between Time Since ?Exit Of Womb/?Fertilisation of Ova and one's appearance that is accurate enough to be used in the way they want. It can depend on genetics, nutrition, and the environment in which the child has lived. (For example, menarch can range from 8 to 16 with an average ~12. That's one helluva spread.)

And the worry is that a technology that might report 1% of rich western kids over 18 when they are under (which is bad enough) might be far more inaccurate for those who've grown up in conflicts zones and are malnourished and stressed from the journey here.

Re: Very difficult for humans

PB90210

Take a picture of Lorraine Kelly and of Sarah Ferguson... which one is older?

Most people would correctly guess Sarah... but the difference is only about 6 weeks!

Re: Very difficult for humans

Phil O'Sophical

It's also extremely culture-specific. Anyone who has travelled abroad much will know that our age-estimation cues are based on subconscious cultural signals. Ask a British man to guess the age of a French or Italian woman, and he'll likely to be way out. If they want to use this to guesstimate the age of travellers at the border they'll need to train it on accurate multicultural data. Good luck with getting that.

Re: Very difficult for humans

Anonymous Coward

Me and my friends used to just go around the back of the tent at Young Farmers events, and slip under the fabric. The Pony Club parties were better though ;)

Skull-measuring

elsergiovolador

The Home Office is spending £1.3m to reinvent skull-measuring. Swap out callipers for a webcam, call it “AI age detection,” and suddenly pseudoscience becomes policy. There is no technology that can tell you someone’s exact age from their face - only rough statistical guesswork, rebranded as accuracy.

But accuracy isn’t the purpose. The point is bureaucratic cover. Once the system is in place, every misclassification becomes “not our fault, the computer said adult.” Children can be pushed into adult detention, adults can slip through as children, and the liability always lands on the black box, never on the Home Office.

It’s digital phrenology with an invoice - expensive, harmful, and designed to move responsibility, not find truth.

I am waiting for Home Office to issue a tender for land on the Moon next.

Re: Skull-measuring

Dan 55

Next up, detecting criminals by measuring eyebrow separation distance calculated by an algorithm running on live video feeds.

Re: Skull-measuring

monty75

Wait til they start cutting people open and counting the rings

Start using non-scientific methods, what could possibly go wrong?

a_foley

”[…] Home Office's approach to assessing children's ages, saying its proposals to use scientific measures to more accurately determine undocumented children's ages would be "ineffective, expensive, and harmful."”

Oh, yes, are you gonna start praying to $DEITY for age verification or what?

Re: Start using non-scientific methods, what could possibly go wrong?

graemep

At least if God does give you an answer it would be an accurate one :)

Not getting an answer is more useful than getting one from any system that does not base it on documents.

What will happen is that some salesman will convince the government they have a solution that will work, lots of money will be spent (a lot more than intended), it will be deployed and not work, then it will be taken down.

How to do govt work:

1. promise whatever is needed to get the same

2. deliver something, even if it is terrible

3. profit!

No AI required

42656e4d203239

Teeth have growth rings... so every inbound person must donate a tooth... this has the added benefit of being able to locate their origin fairly accurately using isotope analysis. What could possibly go wrong?

/icon just in case anyone thinks I am serious...

Hmm I wonder if you could develop a machine than can image a tooth in situ and count the rings using AI? now that might have legs...of course if the putative migrant has no teeth there might be a problem.

Just trust them.

gnasher729

So this person comes and claims they are 13. Goes straight to school appropriate for the age. Benefits go to the guardian who may give them some pocket money. No alcohol until 18. No driving license. Eventually they might get a state pension at age 66 = 53 years from now. “Friends” will be suspect of bring pedophiles. No sex before the age of consent.

However, if they commit a crime they will be treated as adults unless they can prove they are too young. And explain that all very very well if they are suspect of lying.

The mind boggles

Guy de Loimbard

Just at the point I didn't think HMG could be any more inept and ludicrous at procurement, this absolute belter of a request comes out of the Home Office's Orifice!

This, I will wait with popcorn, to see who is going to tender, based on this nonsense request.

Anyone got a link for the actual Home Office request? I would look, but I'm not going to waste what little horsepower I have left today!

Re: The mind boggles

Doctor Syntax

Look at it this way.

HMG keeps coming up with ideas we know are impractical and expecting industry to implement them at industry's own expense or else. HO are prime offenders at this. I keep saying they should provide a proof of concept first.

They're now actually trying this an their (i.e. our) expense. We all know it's going to fail any rigorous trial but AFAIKS such a failure is the only chance of them learning about the reality of what they keep demanding of others. If it works (i.e. if the HO actually learn) then it's a Good Thing.

algorithmic

Rob 63

I've found todays date minus my date of birth is approximately correct

Re: algorithmic

elsergiovolador

This doesn't work for people who were not born on their date of birth. That's why you need AI and usual suspect consultancy to crack this nut.

Beware of a dark-haired man with a loud tie.