Meta Takes Instagram’s AI Age Detection Global to Find Hidden Teens
Meta is expanding AI age estimation to Instagram users worldwide, moving suspected teens into restricted Teen Accounts even if they claim to be adults.
Meta Platforms said on Monday that it is rolling out its artificial intelligence age-estimation technology on Instagram worldwide. The system looks for users it suspects are teenagers, including those who signed up with an adult birthdate, and moves them into the platform’s more restrictive Teen Account settings. The company said the change will place millions more young users under those protections. It comes as regulators and state officials press social media companies to stop relying on the birthdays people type in themselves.
What Meta announced
According to Meta’s announcement, as reported by the technology sites Unite.AI and 9to5Mac, the global rollout builds on earlier deployments in a smaller group of markets. Those included the United States, the United Kingdom, Canada and Australia, and later the European Union and Brazil. “Based on that success, today we’re expanding this technology globally,” Meta said, in a statement quoted by 9to5Mac. Meta said the system can identify teens “even if they use an adult birthday.”
The company said it is continuing to bring the technology to Facebook and plans to launch something similar on Threads, its text-based social app, later this year.
Meta is also starting an awareness campaign for parents. “This month, parents in the United States will begin to see an in-feed video on Instagram explaining our approach,” the company said, according to 9to5Mac. The video links to guidance from Larry Magid, president and CEO of ConnectSafely, a nonprofit internet safety group. The guidance describes a layered approach: keep children under 13 off the platforms, give teens age-appropriate experiences, and help parents make sure their teens list their real age.
How the system works
Meta has said in earlier announcements that its AI draws on several kinds of signals. These include contextual clues, such as birthday messages or mentions of a school grade, and an analysis of photos and videos users have already posted. eWEEK reported in May that the visual analysis looks at general cues such as height and bone structure to estimate a broad age range rather than an exact age. Meta has stressed that the analysis does not identify who a person is. “We want to be clear: this is not facial recognition,” the company said at the time, according to eWEEK.
Accounts flagged as belonging to teens are moved into Teen Accounts. Unite.AI said these accounts come with protections that include:
- Private profiles by default and the strictest settings for messages and sensitive content
- Limits on who can tag or mention the teen, and filters for offensive words in comments and message requests
- A reminder to leave the app after 60 minutes of daily use
- A “sleep mode” that mutes notifications from 10 p.m. to 7 a.m.
Teens under 16 need a parent’s permission to loosen these settings. Meta says users placed in the wrong experience by mistake can change their settings.
Questions about accuracy
Meta has not published data showing how accurate the system is. eWEEK noted in May that the company had released no false-positive or false-negative rates, no breakdown by demographic group, and no figures on how well the tool performs close to the key age lines of 13, 16 and 18. Unite.AI likewise pointed out that the main figure Meta has shared, that more than nine in ten active teen accounts on Instagram kept the restrictive settings, shows how teens use the settings, not how well the AI tells teens from adults.
Independent research suggests the technology is still uneven. eWEEK cited a 2024 evaluation by the U.S. National Institute of Standards and Technology (NIST), the federal agency that tests biometric systems. NIST found that no single age-estimation algorithm clearly beat the others, and that results varied with image quality and with a person’s age, sex and country of birth. The outlet also pointed to a structural problem: a person in a photo is not always the account holder. Australia’s eSafety Commissioner, the country’s online safety regulator, found that 54% of social media users aged 8 to 12 accessed it through a parent’s or carer’s account. 9to5Mac also noted a widely shared case in which a 12-year-old reportedly fooled the system by drawing a fake mustache with an eyebrow pencil.
Why it matters
The expansion follows growing legal pressure on Meta in the United States. In August, the company announced an agreement with 52 bipartisan state attorneys general, which a court approved, Unite.AI reported. Under that deal, Meta committed to invest further in systems that can spot likely under-13 accounts and better recognize 13- to 17-year-olds, even when an adult birthdate was given. Most terms must stay in place for 10 years, and an independent auditor will review compliance every year for five years. Those protections legally apply only in the jurisdictions that signed on. Meta presented the worldwide Instagram rollout as building on those commitments.
Meta is also using the announcement to push its long-running argument that app stores should carry more of the burden. In Australia, the company has begun voluntarily using Apple’s age-range signal, which lets a parent have the device’s app store confirm a teen’s age bracket and pass it to app developers, according to Unite.AI. Meta says that because app stores and developers can choose whether to collect and use that data, protection is inconsistent across apps. It supports legislation that would require app stores to verify age and obtain parental consent, with its own AI acting as a backstop.
The company is also calling on rivals TikTok and YouTube to adopt similar practices. It argues that restricting teens on one service simply pushes them to another. Whether competitors follow, and whether Meta eventually publishes evidence that its system works as intended, will shape how regulators judge the effort.
