Social Media Ban: Age Assurance doesn’t work but the Australian Government doesn’t want you to know that

The Australian Federal Government has today released the full results of their “Age Assurance Trials” which set out to prove to the Tech Companies of the world that they can take on the task of responsibly banning kids under 16 from Social Media, but if you read the headline and quotes from the Government you’d think they have it all solved – in fact they do not. At all.

Here’s how it’s been described, by the Government: “The Albanese Government has today published in full the final report of the independent Age Assurance Technology Trial, which found age assurance technology is effective in protecting young Australians from explicit and age-inappropriate content online.”

The Minister for Communications, Anika Wells, very excited to move forward with the Social Media Ban saying “The Albanese Government is on the side of families and we’re pushing forward with our mission to keep kids safer online through world-leading reforms.  

“This report is the latest piece of evidence showing digital platforms have access to technology to better protect young people from inappropriate content and harm. 

“While there’s no one-size-fits-all solution to age assurance, this trial shows there are many effective options and importantly that user privacy can be safeguarded.”

Here’s the thing – That’s bullshit. Plain and simple.

This comprehensive trial, and it is – awesome, a lot of effort has gone into it, is broken down into several parts. Things like Parental Controls, Age Verification and Age Estimation.

Now, if Parental Controls were the answer (Which they are – but the Government doesn’t want parents to have to be parents in a digital age), then we’d be laughing.

And if Age Verification was the answer, the Government would have a riot on their hands. If you don’t know or understand, Age Verification is the process of taking a Government Issued ID or document and verifying that it’s real and valid. Of course that works, but the Social Media Age Laws clearly specify that platforms (Tiktok, Instagram et al) MUST NOT require a Government ID. This is explicit in the legislation to avoid the drama we’re already seeing online about Government monitoring and control of what we do online.

INSTEAD, it’s reliant upon the “Platforms” to take “Reasonable Steps” to block under 16’s from having accounts. What those “Reasonable Steps” are, with 100 days until the legislation comes into force – we don’t yet know.

SO, we’re back to Age Estimation, and that’ from my reading of the very, very long report – is a shit show.

Yes, it works.

But have a look at this graph:

You see what it shows? How accurately can technology estimate the age of someone? Pretty well when they are under 10, or over 25. But pretty darn poorly if they are between 15 and 21.

The report states that “Age estimation models tend to be most uncertain at or near the threshold (e.g. distinguishing between a 17.8-year-old and an 18.2-year-old), while accuracy improves as you move away from it. As such, the likelihood of a false decision tends to zero outside a reasonable buffer.”

So what then?

• Users confidently above the threshold (e.g. estimated 21+when the cutoff is 18) can be passed with high assurance.
• Users estimated to be near the threshold can be diverted to supplementary assurance methods, such as ID-based verification or parental consent.
• Users confidently below the threshold can be denied access outright.

But wait, there’s more.

I’m not that smart, but from my reading, this graph shows that the average error rate for kids aged 13-15 is 2.74 years. YEARS. So they could be over a year older or a year younger. Try explaining to a 16 year old on their birthday that the technology thinks they are 14 and a half years old. In the end, that parent is going to have to hand over an actual Government ID to stop their child from screaming the joint down because they want access to social media given they just turned 16!

Just to make it clear, the report, when looking at how well Age Estimation could be used in the case of storing digital credentials – states that “Digital credentials that state “Over 13” or “Over 18” offer value in systems that require interoperability and privacy, such as online platforms, e-commerce, and government services, however they are not currently sufficient on their own to support the creation of verified age credentials within digital wallets or holder services for use in regulatory or compliance-critical contexts (e.g. thresholds at 13, 16 or 18).

This limitation is not necessarily a reflection of the underlying model accuracy, but rather of the challenges associated with treating a probabilistic, context-sensitive output (such as “estimated age 18+”) as a reusable, persistent credential that carries the same assurance level as a verified date of birth or government-issued document.“

The report doesn’t highlight these deficiencies, in fact if you go read all the reports their fancy graphic design makes this look like a huge win for the Government’s planned Age restriction and how technology can support it.

I call bullshit on it all.