Meta’s AI age-check goes global: Can it protect teens?

Meta has rolled out artificial intelligence (AI)-powered age-estimation technology worldwide in an effort to better detect teenage users who signed up for its social media platforms with an adult’s birth date.

The world’s largest social media company announced the change on Monday, saying that millions more young people worldwide would be placed in its Teen Accounts following the implementation of the new technology amid concerns that its platforms have failed to protect teenagers.

Teen Accounts have more restrictions and protections designed to provide a safer user experience. They keep the account private by default, limit content types, and prevent strangers from sending direct messages (DMs) to teenage account holders. Parents can also supervise Teen Accounts.

Experts, however, doubt whether the latest step will protect children from online harm.

Here’s what we know about the latest tool and why Meta has deployed it:

What has Meta announced?

Meta, which owns Instagram, Facebook, WhatsApp and Threads, said on Monday that it had started using its AI age-estimation system on Instagram accounts around the world.

“Since first introducing this technology, we’ve been placing more teens in age-appropriate protections and are encouraged by the results. Based on that success, today we’re expanding this technology globally,” the company said.

The system is designed to identify teenagers who signed up with a fake birth date in order to use a regular account. They are then switched to a restricted Teen Account or removed from the platform.

Meta is also rolling out the new age-detection tool for Facebook and plans to implement a similar system worldwide for Threads, another of its social media apps, before the end of the year.

Before Monday, the system had already been rolled out in more than 20 countries. Meta has called on competitors to adopt similar age checks.

How will age checks work, and what protections will they trigger?

Meta says its AI judges whether an account holder is likely to be a teenager, regardless of what birth date they gave when signing up.

The AI searches a user’s account to find “contextual clues” that point to their age. Examples given by Meta include “birthday celebrations or mentions of school grades”.

It also uses “visual cues” in pictures and videos on the account, but Meta says that this does not extend to facial recognition.

Flagged accounts are then moved into Teen Accounts – on Instagram, Facebook and Messenger – which limit how long teenagers can spend on the apps, which content they can access, and who they can message.

In countries that ban social media for users under a certain age, the account may be shut down entirely.

In the United States, Meta will add specific restrictions for under-18s, capping their time on its apps at two hours a day and imposing a midnight-to-6am curfew.

In Australia, the firm says it has started using an Apple feature that verifies a user’s age range when their phone is set up and passes that on to apps they download.

Meta said it supports legislation that would have app stores verify users’ ages, calling it “more effective” than having many developers devising methods to collect the information for each app.

Why is Meta introducing these measures now?

In its announcement, Meta linked the new policy to the outcome of a groundbreaking lawsuit brought by 29 US states and cosigned by 52 US attorneys general.

The company was accused of designing Facebook and Instagram to addict children and collect their personal data, not doing enough to verify users’ ages and even encouraging minors to bypass its existing verification measures.

The suit was settled out of court in August, with Meta agreeing to pay up to $18bn to plaintiffs and to make significant changes to its platform to protect minors, including time limits, nighttime curfews and limits on filters that simulate cosmetic procedures.

Although Meta’s announcement cites the August out-of-court settlement, it has faced two other major legal actions in the US this year relating to the impact of its products on children.

In March, Meta was ordered to pay $375m when a New Mexico court found the company had been misleading about the safety of its platform for minors.

Around the same time, Meta and Google also lost a landmark case in Los Angeles, brought by a woman who said she had become addicted to Instagram and YouTube and faced mental health problems as a result.

Governments have also increasingly intervened to protect minors from what they see as the harms of social media.

In recent years, several countries have introduced social media bans for minors under a certain age, including the United Kingdom, Indonesia, Australia and Turkiye, with many more weighing similar measures.

Is this new system enough?

Sonia Livingstone, a professor at the London School of Economics and director of the Digital Futures for Children Centre, told Al Jazeera that Meta had yet to deliver evidence that its system was preventing underage users from using its platforms.

She pointed to an “absence of clear evidence” that young children could not create accounts or that teenagers who use the platforms were indeed being placed in Teen Accounts.

Livingstone said that Meta’s age check system “poses a privacy risk to all users, not just children, since the point of age assurance is to figure out who the children are”.

Without independent audits to determine how it works, “it is hard to see why we should trust them with our personal data”, she said.

Although Australia banned under-16s from social media in December, its regulator, eSafety, found that more than 80 percent of preteens and young teens were still using it half a year later, largely because age checks failed to stop them.

In that same period, Meta had shut down 462,000 Instagram accounts in compliance with the ban – but eSafety found that more than 69 percent of underage users retained their Instagram accounts.

There is also the risk that teenage users simply migrate to another platform that contains similar risks to their health.

Under the $18bn August settlement, about $5bn of Meta’s payment is conditional on its competitors, including TikTok, YouTube and Snapchat, agreeing to similar measures and payments.

Many of these platforms are also facing regulatory heat for similar issues to Meta. On Friday, European Union regulators threatened to fine TikTok owner ByteDance 6 percent of its global annual revenue for its addictive design.

EU tech commissioner Henna Virkunnen said the platform was “leading users down rabbit holes they cannot easily escape from” and called on TikTok to “step up its efforts to protect users’ wellbeing, including those of minors”.