Why We Built Cova: The AI We Wanted for Our Own Children
Our kids were going to use AI. That part wasn’t up for negotiation. The only question was whether they’d use something built for them, or something built for adults with a child filter bolted on. We looked for the first. We could only find the second. So we built Cova.
We were parents who wanted to let our children use AI. We’d watched them light up asking questions no one had time to answer slowly enough: why stars are different colours, how old the ocean is, whether dragons could have actually existed. The curiosity was real and it was beautiful. And AI, we thought, could be a remarkable tool for that kind of wondering.
Why couldn’t we find that? Because of how the tools that existed were built.
The tools available were built for adults. Some had content filters added on. Some made promises about safety that were hard to verify. None of them felt like they’d been designed around how children think, ask, and feel. They felt like tools borrowed from a different context and retrofitted with varying degrees of care. That gap · between the AI that existed and the AI we’d actually trust · is what Cova is trying to close.
Safe by default is an architecture decision
It’s easy to assume that making AI safe for children is a filtering problem. Block the bad stuff. Add some parental controls. Done.
But filtering is the floor, not the ceiling. A nine-year-old asking “Why doesn’t anyone like me?” might be processing a playground falling-out, or carrying something that deserves far more than a list of friendship tips. An adult AI answers the words · it doesn’t pause to read the moment. A twelve-year-old asking for help with a homework essay should be guided toward their own thinking, not handed finished paragraphs.
Adult AI wasn’t designed around these distinctions. It assumes the person asking is competent, contextually aware, and capable of handling a full answer. That’s not a flaw in children · it’s just childhood. The gap isn’t about missing filters. It’s about foundational design. Safe by default has to be built in from the start, not layered on later.
What parents told us they actually needed
When we talked to parents, the tension was consistent. They saw what AI could unlock: the confidence to ask questions they’d be too embarrassed to raise in class, explanations that finally made a concept click, a space for curiosity to run freely without needing to wait for the right moment. They wanted their children to have that · as a complement to family conversation, not a replacement for it.
In the same breath, they described what kept them up. Conversations they’d never see. Privacy policies they couldn’t decode. The feeling of handing over something powerful without really understanding what it was doing.
What they wanted wasn’t surveillance. They didn’t need to read every conversation about Minecraft or dinosaurs. What they wanted was to know when something mattered · the first time a sensitive topic came up, when the AI encountered something worth flagging, when a pattern was forming. Control without hovering. Visibility without monitoring everything.
That’s what we built Cova around.
Children’s conversations in Cova are never used to train AI models or sold to anyone. The parent dashboard surfaces what matters · sensitive topics, patterns worth noticing, moments worth reviewing · without burying parents in noise. Parents can pause an active session from their own device if they see something in real time. Custom guardrails let parents write rules that fit their family, not just pick from a preset list.
We didn’t build Cova because we think AI is dangerous. We built it because childhood curiosity is valuable · and it deserves better than adult tools designed for a different audience. We started from the right question: what would we actually want for our own children? And then we built it.
Further reading
The Cova team is made up of parents, educators, and technologists who believe children deserve AI that was built for them · not adapted from tools made for adults.