AI Doesn't Remove Humans From the Loop. It Hides Them.

This weekend, my daughter wanted to join a random public Minecraft server.

We had planned to let her play online with a friend, but that friend had temporarily lost iPad privileges. My daughter did not understand why this should change anything.

"Why are you so scared?"

"I can handle it."

"Why do you always worry about bad people?"

It was a classic internet-safety argument. There might be an adult pretending to be a child. Someone might ask an inappropriate question. Another kid might be cruel.

The technology created the opportunity, but the danger was still a person.

Eventually, we turned to Claude in voice mode. We explained both sides, and it did a genuinely great job of helping my daughter understand my concern without making her feel incapable. Then it helped us find a closed, moderated Minecraft club with age verification and a teacher.

She starts this week. She is excited. I am excited for her.

But it struck me: while we were talking about an old model of online safety, we were trusting AI with something much more intimate — our family dynamics. We were asking it to interpret my fear, her frustration, and the complicated space between protecting a child and letting them grow.

AI safety is not simply the next version of internet safety. It is a different kind of problem.

The Old Rules Assume a Person

Most of the internet-safety advice we give children is about other people.

Do not talk to strangers. Do not share personal information. Tell an adult if someone asks you to keep a secret. Remember that people online may not be who they say they are.

These rules assume there is a person making a decision somewhere.

With AI, there may be no person actively trying to hurt a child. No stranger initiating the conversation. No bad actor deciding, response by response, to manipulate them.

The risk can emerge from the system itself.

A child could become attached to something that is always available, endlessly patient, and designed to respond in ways that feel personal. It could affirm when it should challenge, reinforce a distorted belief, or gradually become more compelling than the frustrating, imperfect humans in their life.

That is much harder to explain. I can say, "Some adults online pretend to be kids." That is concrete.

It is harder to say, "This technology can feel like it understands and cares about you, even though it does not understand or care the way a person does. And its responses may be shaped by goals you cannot see."

The Humans Moved Behind the Interface

The humans have not actually disappeared.

People decide what the system remembers. How agreeable it should be. What it recommends. When it challenges you. How hard it works to keep you engaged.

Those decisions then affect millions of people at once — at a scale no individual bad actor could ever reach.

Teach Kids to Be Creators, Not Consumers

Being a creator means understanding that technology did not simply materialize in its current form. Software is a series of choices made by people — and the best way to internalize that is to make some of those choices yourself.

Someone decided what the default would be. Someone chose what success meant. Someone decided which risks mattered and which behaviors the product would encourage.

Children do not need to understand every technical detail. Most adults do not either. But they should understand that the technology they use reflects human decisions, incentives, and blind spots.

I have some faith that my daughter will get this, because she sees me building Hold My Juice. This weekend she asked, "How's Hold My Juice going? Did you add anything new or fix any mistakes?"

I love that framing. In her mind, software is not a magical, fixed thing. It is something people build, change, and sometimes get wrong.

She also participates in programs like RoboHub, where kids go beyond playing with technology and learn to create with it. They make decisions, test ideas, and see that the computer only does what someone designed it to do. Once you have been the someone, you never look at software the same way again.

Tiny Opportunities Everywhere

Recently, I told my kids that Google owns YouTube.

"WHAT? That's crazy."

I asked how that might affect how Google answers our questions. They thought about it and said, "It would probably always show YouTube videos."

Good call.

Seeing It Clearly

I do not want my children to be afraid of AI. The conversation with Claude was useful, thoughtful, and surprisingly lovely.

But I do want them to see it clearly.

AI does not remove humans from the loop.

It just makes them harder to see.