You are not a bystander here.

There is no child sign-up anywhere on this site, and there never will be. You receive an invitation, you create your own account, and you create a Maker profile for your child.

Your child signs in with a pseudonym-based handle you choose and a three-word passphrase Eolari issues. They never need an email address, and Eolari never asks them for one.

Deliberately, almost nothing:

• a pseudonym and a sign-in handle — never their legal name;
• a year and month of birth, so that Eolari can speak to them appropriately and apply the right safety thresholds — never a full date of birth;
• what they make, and the conversations that produced it;
• a record of when AI was involved, and where.

No location, no device identifiers, no advertising profile, no third-party analytics. Eolari has no marketplace and never handles a child’s money.

Consent is separable. You can grant some of these and not others, and change your mind later.

• My child may talk to Eolari’s AI (Mirror and the Forge).
• Eolari may store what my child makes, privately.
• Anonymised, aggregated findings from the pilot may be used to improve Eolari.
• I would like summaries of what my child is working on.

You can see every project your child is working on: their Spark, what they have made, what happened when they tested it, and what they learned. You receive any request that needs your approval, and you are told about any safety event that concerns them.

You do not, by default, get a transcript of every reflective conversation your child has with Mirror. That is a deliberate design decision. A young person working out what they want to make needs somewhere to think aloud, and a child who believes every sentence is being read will stop saying anything real.

If something is genuinely concerning, it is escalated to a human at Eolari and to you — that is what the safety layer is for. Routine reading of a child’s thinking is a different thing, and we have chosen not to build it.

Everything is private by default. Sharing anything more widely — even inside the pilot — requires your explicit approval, every time. There is no public gallery, no feed, and no way for one child to contact another.

• Child-to-child private messaging.
• Followers, likes, or any popularity ranking.
• Public ages, legal names, or precise location.
• Open comments or unmoderated public uploads.
• Behavioural advertising or any algorithmic engagement feed.
• Direct messaging between children and adults.

These are not features we have postponed. They are features that are incompatible with what Eolari is for.

From the guardian portal you can request a full export of everything Eolari holds about your child, or request deletion. You can withdraw consent at any time, which stops AI interaction immediately.

Eolari is a small, closed, consented pilot. The safety architecture described here is real and it is built — but architecture is necessary, not sufficient. Specialist legal, privacy and safeguarding review is part of the path to anything wider, and we would rather say that plainly than imply a completeness we have not yet earned.