Every bar has a bouncer. You know the routine: the squint at your ID, the quick glance up at your face, the nod or the shake of the head. Nobody finds it remarkable. It’s just the cost of entry to a place society decided isn’t for kids.
AI chatbots have quietly grown their own bouncer. As of 2026, Anthropic’s Claude is restricted to users over 18. No under-18s, no exceptions on paper. The door has an age limit now, and it’s a good moment to ask what exactly is behind it.
What the rule actually says
The policy is simple enough that it barely needs explaining: you need to be over 18 to use Claude. That’s it. No tiered kids’ mode, no supervised teen account, no parental dashboard. One line, one threshold.
What’s interesting is the company it keeps. TikTok has also moved its beauty filters behind an 18+ wall, with children’s mental health cited as the reason. Two very different products, two very different companies, arriving at the same number.
That’s not coincidence. It’s a pattern forming across the tech industry, where the answer to “we’re not sure how this affects young people” has become “then young people don’t get it yet.”
Why age gates show up around AI at all
If you’ve never thought hard about it, an age limit on a chatbot can feel odd. A chatbot doesn’t sell alcohol. It doesn’t take bets. It’s a text box.
But a text box that talks back is a different animal from a search engine. Claude and tools like it hold conversations, remember context within a session, give advice, take on personas, and produce answers that sound confident whether or not they’re correct. For an adult with a mental model of how these systems work, that’s manageable. For a fourteen-year-old who has been told this thing is intelligent, it’s a lot murkier.
Compare it to the TikTok filter decision. A beauty filter isn’t dangerous in a chemical-hazard sense. The worry is subtler and cumulative: what it does to a young person’s sense of their own face over months and years. AI conversation raises a similar order of question. What does it do to a developing sense of judgment when a machine is always available, always agreeable-sounding, and never says “I’m just a program, go ask a human”?
Neither company has a tidy answer. The 18+ line is what you put up while you figure it out.
The awkward truth about enforcement
Here is where I’d gently push back on any company claiming this solves something. An age policy is a policy. Most of these gates work on self-declaration, which means the barrier is roughly one click and a willingness to lie.
So the rule does real work in two areas and not much in a third:
- Legal clarity. The company has stated who its product is for, which matters enormously for regulators and liability.
- Signalling. It tells parents, schools, and the public that this is adult software, not a homework buddy.
- Actual prevention. Weak. A determined teenager is not stopped by a checkbox.
None of that makes the rule pointless. It shifts responsibility onto adults, which is arguably where it belongs. If a school is putting AI tools in front of thirteen-year-olds, that’s now a deliberate choice by adults rather than a passive drift into it.
What this means for the AI agent boom
Agents are the current obsession in AI, and they raise the stakes on all of this. A chatbot answers you. An agent goes off and does things: browsing, writing files, sending messages, chaining tasks together with limited supervision. The gap between “gives bad advice” and “takes bad action” is where the real risk sits.
Age restrictions are the earliest, crudest version of a question the whole field will keep facing. Who should be allowed to point a semi-autonomous system at the world, and under what conditions? Eighteen is a blunt answer to a question that probably deserves a more careful one, but blunt answers arrive faster than careful ones.
Where this leaves the rest of us
If you’re an adult using Claude, nothing changes for you. If you’re a parent, teacher, or anyone with teenagers nearby, the useful takeaway isn’t “AI is banned for kids.” It’s that the companies building these tools are openly unsure about their effect on young minds, and they’ve said so through their terms of service.
That’s more honest than the industry usually manages. A bouncer at the door isn’t a solution to anything happening inside the bar. But it does tell you the owners think something in there needs watching.
🕒 Published: