Hasbro built Sixth Wall. Here's the playbook for every other IP owner.

Hasbro just put a price tag on personality. On June 3 it announced Sixth Wall, an in-house AI studio built to sell something the toy business has never sold before: behavioral licensing. Not the right to print a character's face on a lunchbox. The right to its behavior: what it says, how it plays, the way it answers a six-year-old's question about the moon.

The announcement names two load-bearing pieces. CharacterOS, the studio's system for carrying a character's behavior across products and partners. And a voice partnership with ElevenLabs, so licensed characters sound like themselves when they speak. The Hollywood Reporter read the move plainly: Hasbro intends to license characters as beings you talk to, not just images you look at.

If you own characters (an animation catalog, a publisher's backlist, a mascot three generations grew up with), you likely read that news and felt a clock start. Good instinct. Here is what actually happened, why it rewrites the licensing business underneath you, and what your options now are.

What Hasbro actually announced

Strip the commentary and the confirmed facts are short. Sixth Wall is a new AI studio inside Hasbro, announced June 3, 2026. Its stated business is behavioral licensing: granting partners rights to how a character behaves and converses, the way image licensing grants rights to how it looks. CharacterOS is the system built to carry those behaviors. ElevenLabs supplies the voices. That's the press release.

Notice what the announcement does not include: a rate card, standard terms, settled norms. Those are unformed. Everything past the facts is inference, ours included, but the inference is not difficult. One of the largest character owners on earth has decided that behavior is a separate, sellable layer of IP, and has stood up a studio to package it. Toy giants do not build studios for hypotheses. They build them for product lines.

Conversation is a new rights surface

Character licensing has priced one thing for roughly a century: the image. The whole apparatus of style guides, approval rounds and category royalties exists to control how a character looks when somebody else manufactures it. The system works because images hold still. Approve the artwork once and every unit ships the same face.

Conversation does not hold still. A talking character produces new sentences every day, one child at a time, unsupervised. The contract that covers a lunchbox has no clause for that, so a new set of questions lands on the table, and none of them have standard answers yet:

• What may the character say, and what must she refuse to discuss?
• Who writes her answers to the hard questions a child will eventually ask: death, fear, divorce?
• Who owns the transcripts? A year of one child's conversations is an asset no image deal ever contemplated.
• Whose safety stack sits between the model and the child, and who faces the regulator when it fails?
• What happens to the voice, the memory and the data when the contract ends?

The commercial pull explains the timing. Children already expect characters to talk back, and analysts tracking the smart toys market project double-digit growth through 2030. Conversation was always coming to the toy aisle. The open question was who would own the rails.

The squeeze on everyone who isn't Hasbro

Hasbro can afford to build CharacterOS because it can spread the cost across one of the deepest character vaults in the world. Most IP owners cannot, which leaves exactly three routes. We build the third one, so here is our bias stated plainly: we are the neutral-OS vendor making the neutral-OS argument. Judge the logic, not the seat.

Route one: build your own

You won't, and that is arithmetic, not an insult. A CharacterOS of your own means speech engineers, a child-safe model stack, a parent app, voice infrastructure and a compliance machine, maintained forever. Spread across a vault the size of Hasbro's, that cost amortizes. Spread across five beloved characters, it doesn't. The math does not improve with enthusiasm.

Route two: license into a walled garden

Bring your character to a giant's platform and the demo will be wonderful. But your character now lives on their terms: their pricing, their data policies, their roadmap, and a shelf where their own characters come first. Image licensing already taught this lesson once. Distribution power becomes pricing power, and it never flows toward the licensor.

Route three: ship on a neutral OS

Keep the character. License nothing away. Run its behavior on an operating system that owns no competing catalog, inside a plush or shell you already manufacture, under your own brand. This is the Android argument. Apple built the closed, vertically integrated stack, and every phone maker that wasn't Apple consolidated on the open alternative, which ended up in most of the world's pockets. Hasbro just built the iOS of character AI. The rest of the market needs an Android.

Your character on PlayOS, concretely

Here is the shape of the deal, with the numbers we publish. PlayOS is free to adopt: voice engine, safety filters, OTA updates, parent app, SDK. The hardware is the Kheelona Magic Box, a companion module at $10–50 per unit that drops into your own plush or shell. Integration runs 4–6 weeks from first call to shelf-ready, you set the retail price and keep the hardware margin, and the engine underneath is our own sub-1B Voice SLM, opening to API credits from Q4 2026.

• Character config. You define the personality, the vocabulary and the refusal list. She behaves as written, on every unit, through every update.
• 10+ languages. Hindi, English and regional Indian languages. A character that speaks a child's mother tongue is a different product from one that doesn't.
• Your IP stays 100% yours. We are the OS inside, not a co-owner of your character or your world.
• Safety you inherit. COPPA and GDPR-K compliance and age-graded responses are built into the OS, so your deal does not begin with a safety hiring plan. The safety stack is public.
• Revenue after the sale. Custom stories and character tie-ups ship as subscriptions, revenue-shared with you. The character keeps earning after the shelf.

We run our own character on this exact stack. Lumi, our companion plush, is live with 10 families today; her personality, refusals and languages are PlayOS configuration, nothing more. The kids choose her over screens. That sentence is the whole pitch.

A rights checklist for any character-AI deal

Take this list into any negotiation, including one with us. If a clause is missing, the deal isn't done.

• Voice rights. Who may synthesize the character's voice, trained on which recordings, stored where, revoked how? A voice model is a copy of a performer's instrument. Treat it like one.
• Conversational boundaries. The refusal list belongs in the contract, not the marketing deck: which topics, which age bands, and who may change them after launch.
• Transcript ownership. A child's conversations are regulated data and a training asset at the same time. Settle who holds them, who may learn from them, and how a parent deletes them. Silence here favors the platform.
• Kill-switch. When the character says something off-model at 2 a.m., you need an OTA path that corrects every unit in hours. Ask to see it demonstrated, not described.
• Contract end. The voice stops, the memories are deleted or handed back, and the character walks away clean. The end of a deal should never be the end of the character.

Sixth Wall settled the argument about whether character behavior is a rights business. It is one now. What is not settled is whose terms your character will live on, and those defaults are being written this year, in contracts like the ones above. Move while the ink is wet.