Giving digital youth a voice in AI — and the safety harness behind it

AI is going to be one of the most personal technologies digital youth ever use.

They will talk to it, learn with it, build with it, confide in it. And voice — a child's voice, a teenager's voice, in real time, into a system that listens, remembers, and replies — is where this gets most personal of all.

A minor's voice carries data no keyboard does. A real-time voice agent isn't a tool. It's a presence. And that changes what has to get built underneath it.

Right now, most teams shipping AI to youth audiences have two bad options: lock under-18s out entirely, or onboard them in ways their parents and the law never agreed to. The first cedes the audience. The second is a lawsuit waiting to happen — and worse, a product quietly doing the wrong thing to the users it was meant to serve.

Digital youth should get the upside of AI — the patient tutor, the storytelling collaborator, the language partner, the AI character they build a world with. The bar for whether they get to do that shouldn't be "we couldn't figure out the legal side." It should only ever be: is this right for them?

That's the question every team building AI for youth should be answering. And it’s hard to answer.

By the day, by the country

What "done right" means changes by the country.

In the US, the 2026 COPPA Amendment treats AI training on a child's data as a separate parental consent, not buried in the terms of service.

Brazil's ECA Digital requires verified age 18 for AI-driven profiling and targeted ads, and parental controls that identify the adults the child or adolescent is interacting with.

Korea fined the maker of the "Iruda" chatbot for collecting data on 200,000+ children under 14 — and Korea's threshold is under-14, not under-13, a quiet trap for anyone built against a US baseline.

Australia's eSafety Commissioner has moved past rule-writing and started naming names, issuing transparency notices to the providers of major AI companion chatbots.

The UK's AADC, the EU AI Act, GDPR Article 22, and the Digital Services Act each add their own layer. All of it moving faster than the product cycles of the teams building inside it.

What k-ID does

k-ID models that entire gradient and turns it into a single compliance and safety harness any youth product can ship on.

Take a single signal — a child's or teenager's voice, going into an AI tutor. Speaking the question is one decision. That voice being retained so the system can build a profile of them over time is a second. That voice being used to train the next model is a third. Each is its own consent, with its own jurisdictional default, governed by its own law. Most AI products collapse all three into the privacy policy or terms of service and call it consent.

By extending k-ID’s CDK to address these parameters, each one is treated correctly, per market.

And the system sitting on top is age-aware in real time — it knows who it's talking to, what's permitted for that user in their jurisdiction, what's currently enabled at the family level — and the experience adapts continuously, without anyone redeploying anything.

With Inworld AI

Inworld powers voice and character behind some of the most ambitious AI experiences in market.

Together, we're closing the gap between shipping AI to youth audiences and shipping it to them safely — so any team building on Inworld, like the just announced Realtime TTS-2, can lean on the full k-ID CDK for AI underneath their work.

An age gate tuned to each jurisdiction. A verifiable parental consent flow that stands up to regulator scrutiny. The right defaults set on every AI decision a youth product has to make, country by country. 

One integration path. Every market. Every age band. Parents in the loop, in real time.

Building the voice

Somewhere right now, an eight-year-old is picking up a microphone and asking an AI a question. So is a fourteen-year-old. They deserve better than a 13+ wall they'll just climb over. They deserve a system that knows what's right — and that adapts the moment any of it changes.

That's what we're building: a voice for digital youth in AI. Starting with the literal one.

— Kieran