Only social media companies know for sure how the teen ban will work

The TV cliffhanger has been a dying art for almost a decade now, but Australia's teen social media ban is single-handedly reviving it.

Questions as to how the saga will end have been piling up since episode one last November, when parliament passed a law to purge under-16s from major platforms such as Instagram, TikTok and Snapchat.

Like any good TV drama, it's produced its share of intrigue — which platforms will actually be covered? What's a VPN? And how do the Wiggles feel about it?

The reply all along has been "stay tuned", and those who've done so have been drip-fed some of the answers, to varying levels of completion, in a drawn-out series of nail-biting instalments.

Which platforms will be covered by the ban? We still don't have an exhaustive list, but it's most of the big ones you can name, including YouTube!

Sluggish mid-season ratings perked up with the bombshell reveal that age-checking technology isn't always as accurate as hoped — especially if you're close to 16 years old and not a white male.

But with less than three months to go, the season's biggest mystery remains unsolved: how will social media companies actually check who's over 16?

The latest cliffhanger

Many loyal viewers had high expectations that the answer would become clear in this week's episode, with eSafety Commissioner Julie Inman Grant unveiling long-awaited regulatory guidelines, but there was no big reveal in the end.

Rather than giving a concrete list of acceptable age-checking methods, Inman Grant announced a "principles-based approach", stating that age-checking systems should be "reliable, accurate, robust and effective".

How exactly the platforms go about the messy business of testing their users' ages — scanning faces, checking IDs, using third-party data from an app store, or making an educated guess based on users' online activity, to name a few options — will be up to the companies themselves.

Addressing the media on Tuesday, the commissioner seemed aware her audience might have been expecting something less vague.

"It may be a surprise to some that the guidance does not mandate a single technology approach … or set a required accuracy rating for age estimation," Inman Grant acknowledged.

What did emerge, though, was a clear expectation that social media companies provide Australians with more than one way to demonstrate their age — a so-called "waterfall approach", where ID is never the only option.

And apparently, no-one is expecting perfection.

"We don't expect that every under-16 account is going to disappear on December 10," Inman Grant said.

"What we will be looking at is systemic failures."

The exact definition of "systemic failures" is yet another mystery.

It's unclear just how "reliable, accurate, robust and effective" age-checking methods will need to be in order to avoid attracting a $50 million fine, with no firm thresholds set out.

In any case, eSafety will be marking homework.

The regulator will have the power to force companies to hand over internal data as part of compliance checks, such as reports on how many young people they've kicked off their platform.

The big reveal? An educated guess may be good enough in many cases

If there was a big reveal in this episode, it might have been that social media companies won't need to ask everyone's ages — in many cases, an educated guess will do.

In technical terms, an educated guess as to a person's age is known as "age inference", and the eSafety Commissioner's new guidelines suggest it may figure much more prominently in the final shape of the policy than anyone initially imagined.

"Existing data may be sufficient to satisfy providers that the account holder is over the age of 16," the guidance states.

At the more sophisticated end of the spectrum, age inference can be remarkably accurate, drawing on a range of behavioural signals and contextual data, rather than face scanning or government-issued ID.

An age inference system might consider how long an account has existed, the interests and personal networks of the user, the content of their posts, or even analyse their voice and use of language.

Put simply, if all your friends are 13, you speak and write like a 13-year-old, and your online activity drops off during school hours, a social media platform might ask you to prove your age.

On the other hand, if you've had an account since 2007, your friends are all in their 40s and you're a member of several online parenting communities, you might not ever hear from them.

A key appeal of age inference, according to eSafety, is that no extra data is required from the user, making the whole experience "frictionless".

Granted, it's probably not the best way to tell a 15-year-old from a 17-year-old, but it might be enough to keep the majority of Australian adults from even noticing that the social media ban has come into force.

There's a certain irony to the fact that, amidst all the anxious guessing brought on by the shortage of policy detail over the last 10 months, very few people guessed at this ending.

Of course, we can't know how heavily social media companies will lean on this option, but we do know it's already a common in-house practice for many of them.

It's also safe to assume companies will want to avoid any unnecessary headaches for themselves or users come December, and with eSafety's conditional green light, age inference will be all the more enticing now.

Speaking alongside Inman Grant this week, Communications Minister Anika Wells said platforms "know an awful lot about us" already, and millions of Australians could be about to learn exactly how much.

But until the season finale rolls around on December 10, it's all an educated guess.