Social media ban for teens: Tech expert Tony Allen tasked with making Government plan work

Australia’s world-leading move to ban under-16s from social media was one of the biggest political decisions of 2024, but it left one crucial question unanswered – how to enforce it?

Tony Allen, head of the UK-based Age Check Certification Scheme, has been tasked in the coming year with finding a solution, sifting through some 50 different existing approaches to gatekeeping the internet around the globe to land on the best fit for Australia’s needs.

In an interview with The Nightly, Mr Allen revealed his company would trial a complex mix of three types of technology, but he cautioned that there was “no silver bullet” to easily exclude children nationwide from a diverse range of social media sites.

After the current evaluation design phase, trials will be conducted in the first half of 2025 to find the right combination of age verification, age estimation and age inference tests that likely be used when the law comes into force, in a process called “successive validation,” he said.

After testing a variety of possible blocking tools on about 1,100 Australians, the company must report its recommendations back to the Government by June, before social media companies form their own strategies on how to enact the ban when it hits in 2026.

It’s a timeline that Mr Allen admits is “definitely tight.”

Age verification

Communications Minister Michele Rowland has already ruled out mandating social media users to upload proof of identity directly to social media platforms.

But age verification, where a date of birth is established through digital IDs like passports or driving licences, remains part of the blueprint being studied by Mr Allen’s team, who will test differences between fake and legitimate documents.

He explained, however, that technology to verify someone’s date of birth did not necessarily require the uploading of documents, citing banking systems that can release this date without providing other information.

Mr Allen stressed privacy was central to the debate on how to control minors’ internet access, not only when it came to documentation but also the limits on tracking individuals’ digital footprint.

“You would want systems that are built with that privacy preservation in place. You want top trust in them. You want them to be certified. You want them to be accredited,” he said.

Ultimately, determining the parameters of privacy lay with the eSafety Commissioner, the Australian Government’s independent online safety regulator, he said, but there were a number of precedents around the world where age verification tests were already used as an access point.

Online gambling, the purchase of potentially dangerous items like knives, or activities involving financial transactions, were examples where more privacy intrusion was required of users, he said.

Age estimation

Age estimation is a more nuanced way of telling someone’s age, based on features that change as we mature.

“Our faces get older, our voices get deeper, our hands change the way they behave,” said Mr Allen. Hair grays, eyebrows begin to sag, and AI can be trained to recognise the indicators.

“You can analyse it in the same way as you would when you’re walking into an off license and the person behind the counter looks you up and down and goes, ‘yep, they’re old enough!’” he said, referring to a British term for a bottle shop.

Current technology can estimate age within about one and a half years. It does not try to work out who you are but compares your features with the training data gathered from about 100,000 people.

Age inference

Age inference is when you deduce someone’s age from factors other than a precise date of birth.

“Somebody wearing a primary school uniform, you could infer that they’re more than likely going to be a primary school child,” said Mr Allen.

“Somebody who is a commercial airline pilot - you can infer that that means they’re going to be over 21 as that’s the minimum age to be an airline pilot.”

Inference clues could be found in things like email addresses, firearms or credit card ownership, or whether you have a mortgage.

“There are loads of facts that you can get out there about people from which it’s a reasonable inference to say that person is likely to be an adult or a child,” he said.

Successive validation

Combining the various methods together is called “successive validation.”

Referring back to the bottle shop analogy, Mr Allen equated it with the person behind the counter not thinking a customer looks old enough to buy alcohol, prompting them to ask for ID, and then for further proof of date of birth.

“They all add together, and you’ve got permutations and combinations alongside age verification,” he said.

Systems that work “provide a range of options that you can pick from that build up a picture to get a level of confidence,” he said.

“It’s very much a case of understanding the objective, which is typically stopping children from accessing the material and picking the right solution for the case that you have, what your customers want and need.”

Preventing loopholes that inventive young minds could exploit to get round the ban will also be part of the trialling process.

If, for example, an older sibling helped an underage user to access social media, “there’s only a certain degree to which you can stop that from happening,” said Mr Allen.

But he suggested that access would be short-lived, as child safe monitoring algorithms would pick up anomalies that would require further verification.

Ultimately, Mr Allen’s company has been tasked with examining whether the technology works.

Building trust and confidence in the system is the next stage of the Government’s mammoth task ahead.

And as Australia steps out with a global-first experiment, it can be assured that the eyes of the world will be watching.