Meta* rolls out AI age checks on Instagram* worldwide: children under 13 will lose access

Anyone who has built a signup flow for a social network or an age-restricted service knows the problem: a “date of birth” field verifies nothing. A child enters 1990 and gains access to a product intended for adults. Meta has decided to close this loophole with AI and, according to The Verge, has begun rolling out age checks on Instagram worldwide.

What exactly happened

The company is launching a system that estimates a user’s age with an artificial intelligence model, rather than relying on the age they entered themselves. What happens next is straightforward, with two possible paths.

The first concerns teens aged 13 to 17. Meta* expects stronger safety settings to be enabled on millions of their accounts. If a teenager has been posing as an adult, the system is supposed to detect that and move them into a more protected mode.

The second path is stricter: accounts belonging to children under 13 will be blocked. This is not about “limiting features,” but removing access altogether, because the minimum age to sign up for Instagram* is 13.

The original report provides no technical details: the signals the model uses, its accuracy and the appeals process are all unknown. Those questions remain open.

Why this matters beyond Meta

The main shift is in the approach, rather than the model itself. Age is no longer information a user enters once; it becomes an inferred attribute that the platform checks independently and reassesses over time.

Consider a simple example. A user says they are 25 when signing up. A month later, the system sees that the account behaves and looks like a schoolchild’s profile. Previously, the platform had no reason to question the stated age. Now it can switch the account to teen mode without waiting for a report.

For product managers and founders, this has three practical implications:

• If your service caters to teenagers or offers content rated 13+, 16+ or 18+, age controls will need to be designed into its architecture, rather than treated as a checkbox.
• Regulators in different countries are increasingly demanding meaningful protection for minors, and major platforms are setting the standard that enforcement authorities will look to.
• You need a plan for mistakes: what happens to an adult whom the model mistakes for a teenager, and how quickly can they challenge that decision?

Limitations and risks

AI age estimation is always probabilistic. It produces two types of errors: false positives, when an adult is classified as a child, and false negatives, when a child goes undetected. Their costs differ. The first damages the user experience and trust; the second undermines safety and the platform’s reputation.

There is also the question of privacy. To estimate age, the system needs to analyze account data, and users are entitled to know exactly which data it uses. Meta* has not yet disclosed those details in connection with this announcement, so it is too early to assess the balance between protection and surveillance.

Finally, there is an arms-race dynamic. Teens quickly learn to bypass restrictions: they create new accounts, change their behavior and use other people’s information. Any age-checking system will require continuous fine-tuning.

What to do in practice

If you are building a product where age determines access to features, there are three things worth doing now. First, store declared age and estimated age in two separate fields. Second, work out which features change when an account switches to teen mode. Third, build in an appeals process from the outset, because mistakes are inevitable.

Meta’s* move shows where things are heading: age checks are moving out of the signup form and into a model running in the background. Anyone serving teenage users would do well to prepare before a regulator requires it.

• Meta and Instagram are designated as extremist organizations and banned in Russia.