It could take just 40 seconds: that’s how long Australians might have to wait for age assurance technology to assess they’re allowed to be on social media from December, when the under-16s ban comes into effect.
But the reality is, for many users, it could be a much more complicated and time-consuming process.
The official trial of age assurance technology, whose report was published on Sunday, found that facial age estimation tests – which estimate a user’s age using an image of their face – took on average 40 seconds. But in the datasets the trial has released, it took some users between 10 minutes and an hour to get through the process.
That’s before taking into account whether the system has correctly estimated their age.
The federal government and the trial report downplayed concerns about error rates in facial age estimation, but the report acknowledged there will be errors, particularly for users closer to the 16-year-old age limit.
Yoti – one of the more mature age assurance technology providers tested in the trial – had results showing an error rate of 34% for 14-year-olds and 73% for 15-year-olds, where the system incorrectly said those users were 16.
For a social media ban aimed at keeping 13- to 15-year-olds off social media, that’s a massive cohort that could potentially escape the planned ban.
Sign up: AU Breaking News email
The data reveals some of the age estimations can vary widely, particularly if you are non-Caucasian, female or older, according to the report. The data reveals one 12-year-old was estimated to be 85.
The report conceded that errors in age estimation technology were inevitable and fallback options would be required for platforms where the user says they’re not the age the technology guesses they are.
That would probably lead anyone wrongly assessed as being under the age limit to have to provide IDs such as passports to be able to get on to social media – a much more privacy-intrusive method of verifying age.
Worryingly, the report found some of the providers offering age verification tools were preempting expectations from law enforcement that the identification data might need to be retained for other purposes, which the report noted could risk privacy breaches.
Concerns around wrongful bans
What if someone doesn’t have an ID or can’t get back on to the platforms despite the additional hurdles?
The communications minister, Anika Wells, said the government was mindful that wrongful bans could be an issue, and under the rules there should be a way for people to appeal against a decision and that should happen quickly.
after newsletter promotion
However, any reporter working in the technology space in the past few months would know that Meta, in particular, is often not responsive to users requesting appeals when their accounts are banned. Whether the company develops a sudden sense of urgency for these accounts remains to be seen.
Tellingly, in media interviews on the report on Monday, Wells talked up what is referred to in the report as “age inference”. This is the most frictionless and least privacy-invasive method.
There is a view that social media companies collect vast amounts of data on their users already and, through this data, they can figure out the age of each user. A user, for example, who joined Facebook in the Kevin 07 era would be well past the age cut-off.
There are also behavioural signs. Wells used the example of interacting on social media with retirees about caravanning being an indicator someone is older, and talking about Kpop Demon Hunters with 13-year-olds being a sign someone is a teenager.
In an ideal world, in around 100 days when the ban comes into effect, the platforms will use this method for the vast majority of adults and no one will notice. That may be the case for some platforms – and we don’t know which method any of them will use yet – but smaller providers or those that deliberately avoid collecting much information on their users will have to employ some of the other methods.
Some smaller providers may choose to opt out of Australia entirely. Twitter-X alternative Bluesky blocked users in Mississippi last month after a similar age assurance law passed in the US state, citing costs in developing age verification systems and lack of resources.
Bluesky has not said what the plan is for Australia yet, but it’s not difficult to predict some smaller companies will choose to opt out of Australia altogether rather than face either the cost of compliance or fines of up to $50m.
