Is an AI companion actually safe for your child? Here is what to look at

A calm, specific look at what Lumi stores, what parents control, and where any AI product's limits are, including ours.

What Lumi stores, and what it does not

Lumi collects only what it needs to personalise conversations and run the service: your own parent account details from Google sign-in, your child's profile (name or nickname, age, language comfort, interests, and similar), their XP and badges, and the text of conversations, so you can read them back in Parent Corner.

Lumi does not store audio recordings of your child's voice. Speech-to-text runs in your device's browser, and only the resulting text is sent onward, to generate a reply and save the transcript. That reply text is sent to a cloud text-to-speech provider to generate spoken audio, which again never involves a recording of your child.

Lumi does not sell personal data, does not show advertising to children, and does not use children's data for behavioural tracking or profiling. Deleting a child's profile removes their profile and transcripts from the active database, and you can do this yourself, or by contacting us.

Parent supervision and controls

Lumi is set up by a parent, not by a child. You sign in, create your child's profile yourself, and everything that follows sits under your account. Parent Corner is where that supervision becomes concrete: you can read every past conversation, see new words your child has picked up, block specific topics, set quiet hours, and set a daily time limit, all enforced by Lumi rather than left to an honour system on the device.

How conversations are shaped for children

Lumi is built to stay strictly a play companion, not a homework helper or a general-purpose assistant, and it refuses requests outside that role. It never asks a child for their last name, home address, school name, phone number, or any way to be found or contacted. There is no romantic, scary, violent, or mature content. Every reply passes an automated safety check before it is spoken, and if a child says something that suggests they are hurt, scared, or unsafe, Lumi is designed to gently point them toward a trusted adult rather than continue as if nothing happened.

What to watch for in any AI product

None of this is worth checking only for Lumi. Any AI product a child talks to is worth the same plain questions: What actually gets stored, and for how long? Is there a real person, a parent, who can see what the child is saying? Are limits like time and topics enforced somewhere a child cannot simply switch off, or do they live only in a setting a curious 8-year-old could undo? Does the product say plainly what it is not built to do, or does it stay vague about its own limits?

A calm way to check any of this, for any app, is to actually open its privacy policy and read what it says it collects and shares, rather than going by how the app feels to use.

Where the limits are

Lumi's replies are generated by artificial intelligence and pass through an automated safety check, but no automated system is perfect. Lumi is a play-and-practice companion. It is not a substitute for professional educational assessment, speech therapy, medical or mental-health support, and it is not a substitute for your own attention as a parent. It is designed to work alongside supervision, not instead of it, and it is worth checking in on your child's conversations in Parent Corner every so often, rather than treating any AI product as something you set up once and never look at again.

Questions parents ask

Does Lumi store recordings of my child's voice?

No. Speech-to-text runs in the browser, and only the resulting text moves onward to generate a reply and save the transcript. No audio recording of your child is stored.

Can I see what my child says to Lumi?

Yes. Parent Corner shows you past conversations, along with new words your child has picked up.

Does Lumi show ads or sell data?

No. Lumi does not sell personal data, does not show advertising to children, and does not use children's data for behavioural tracking or profiling.

What happens if my child says something concerning to Lumi?

Lumi is designed to gently steer a child toward telling a trusted adult if they say something that suggests they are hurt, scared, or unsafe. Every reply also passes an automated safety check before it is spoken.

Is an AI companion a replacement for parental supervision?

No, and Lumi is not built to be one. It is set up and supervised by a parent, with full visibility into conversations, and it is not a substitute for professional support or your own attention.