Meta Shows Teens the Door: Its AI Characters Are Going Back to Class

Meta has paused teen access to its AI characters across all of its apps, including Instagram and WhatsApp, and the decision applies worldwide. The company stresses that it isn't abandoning the product, just pausing it: Meta is preparing an updated version of its AI characters for teens that will account for users' age. Adult Meta users can still message virtual versions of celebrities such as Paris Hilton or Tom Brady.

What changed for teens

Meta has officially confirmed that it is temporarily blocking minors from its AI characters across all its apps, including Instagram and WhatsApp. The restriction applies globally. For adult users, the service works as before: they can, for example, message virtual versions of Paris Hilton or Tom Brady. According to the company, this isn't the end of the story but a pause to build a more suitable version of the product, one that accounts for users' age.

How Meta's AI characters came about and what went wrong

From the start, Meta's AI characters looked like an attempt to catch up with Character.ai. The idea was to let users chat with someone who looks and talks like their idol. But the algorithms behind these bots sometimes produced content that didn't fit a family app: a virtual companion could discuss topics not meant for a schoolkid or come across as overly pushy. The root cause is that large language models are inherently prone to hallucinations and don't always respect the filters placed on them.

Why Meta paused access now

Regulators around the world, especially in Europe and the US, are closely watching how tech companies protect children. This comes against the backdrop of years-long disputes over Instagram's impact on teens' self-esteem, and the company can't afford another high-profile AI scandal. It's easier to pause the service now and rebuild the model than to pay billion-dollar fines later and explain itself before the Senate.

What challenges remain unsolved

Meta hasn't yet specified what exactly will change in the updated teen version. Standard reinforcement learning from human feedback doesn't always cope with teen slang, context, and attempts to get around the system with jailbreak prompts. Building an AI that is both charismatic and completely harmless is a hard problem, and the company has acknowledged that its current work falls short of that standard, choosing not to risk its reputation for a short-term bump in metrics.

At the same time, an overly strict version risks turning out bland: if teens lose interest in talking to the bots, they may move to other, less controlled platforms. If even Meta, with all its moderation resources, admits it can't handle this task in the current format, startups in this space will have an even harder time.

What exactly did Meta pause?

Meta has paused teen access to its AI characters across all its apps, including Instagram and WhatsApp. The restriction applies worldwide, and for adult users the service continues to work as before.

Has Meta abandoned the idea of AI characters for teens?

No. The company said this isn't abandoning the product but a pause: Meta is developing an updated version of its AI characters for teens that will account for users' age.

Why did Meta take this step?

The algorithms behind the AI characters sometimes produced content unsuitable for a family app, and large language models are inherently prone to hallucinations and don't always respect the filters placed on them. This comes amid close scrutiny from regulators in Europe and the US over how tech companies protect children, and years-long disputes over Instagram's impact on teens' self-esteem.

*Meta is designated an extremist organization and banned in Russia.