Every parent in Israel today feels some version of the same thing, even if they have never put it into words: something enormous is happening, it is going to shape our children’s lives, and we have very little idea what to do about it. That is understandable. Parents themselves are still processing the AI revolution and trying to navigate the rapid changes it is bringing.
It is no longer enough to learn how to use the technology and all the tools surrounding it. Parents also have to figure out how to introduce those tools to their children intelligently and safely, while protecting both their privacy and their mental health.
And while adults worry about what AI might do to the cognition of generations raised alongside it, children are already there. They entered this world faster than many parents expected, and without waiting for us to decide whether they should.
So what now? How do parents help children navigate this world, encouraging them with one hand while protecting them with the other?
When should children enter the world of AI?
So when should AI use begin at home? Common Sense Media, which has rated content for children for two decades, updated its guidance this month. For children under 5, it recommends avoiding generative AI. Between ages 6 and 12, use should be limited and always accompanied by an adult. From 13 onward, children can begin using it more independently, but with supervision.
And there is one hard rule: until age 18, AI should not be treated as a source of emotional support. The major tools, including Claude, Gemini and ChatGPT, work well in Hebrew and are improving constantly. Claude is available only to users 18 and older, while ChatGPT and Gemini offer more protected accounts for teenagers 13 and up. Gemini also offers a supervised option for younger children that parents can manage through Family Link.
None of this makes any of the tools “safe for children.” Some companies are already trying to estimate whether a user is a minor even when the age entered on the account is false, but these systems are imperfect, their restrictions can often be bypassed and the rules are changing quickly.
Just as children learned to lie about their age during the early internet era, they can do the same here. According to Bezeq’s annual internet report, conducted by Kantar and published in December, the average Israeli child begins using AI at age 10.
A separate study by the Israel Internet Association and the nonprofit ELEM, presented to the Knesset Science and Technology Committee in May, found that 96% of teenagers use AI chatbots, with 85% using them every day or almost every day. Not most of them. Almost all of them.
One figure in the Bezeq report deserves particular attention: 61% of teenagers said they felt AI was replacing the need to consult their parents. That is frightening. At that point, this is no longer simply a technology story. It becomes a question of who is answering your child’s questions about life.
AI is already part of children’s lives. For parents, the question is no longer whether to allow it, but how to deal with it from here.
(Photo: Shutterstock)
First, build skills
Start with the side of the issue almost no one talks about: how to use this technology to push a child forward. The clue is in the name. This is a revolution in generative AI, and generation is precisely the capability children should learn to exploit.
Do not teach them only to ask it questions. Teach them to make things with it. That can mean images, videos, jokes, stories, websites and games. A 10-year-old can spend an afternoon building a simple computer game, making a birthday video for a grandparent or creating a small website for a class trip, without knowing a single line of code.
The learning happens through the loop. The child describes something, the machine produces it, the child looks at the result and says, “No, make the enemy faster,” and watches the product change. Describe. Look. Correct. That is structured thinking, even if no one calls it a lesson.
It also offers an answer to one of parents’ biggest fears. A child who simply receives an answer can avoid making an effort. A child who creates something must know what they want, judge what the system produced and correct it. The machine cannot do those three things for them.
With younger children, this can work beautifully as a continuing bedtime story in which the child is the director rather than the listener. They decide who enters the next chapter, what happens to the hero and how the story ends.
With a teenager, let them catch the machine making a mistake in an area they know well. Ask it about their favorite sports team or a TV series they know by heart.
The confidence with which AI can state something completely wrong may be one of the most important lessons they can learn about the technology, and it is better for them to discover that themselves than simply hear it from a parent.
Parents should also use AI in front of their children for something real, such as drafting a letter to a teacher or planning the summer vacation. When a child sees you stop and say, “Wait, that doesn’t sound right. Let’s check,” they learn the attitude rather than merely the tool.
(Photo: Getty Images)
And let your children teach you. The American Academy of Pediatrics recommends asking two questions, with the second perhaps the more powerful one: “How are you using AI?” and then, “Show me how it works.” That turns the child into the teacher and the parent into the student.
Children should also understand, at least broadly, what these systems actually are. An AI language model is a statistical system predicting what comes next. It has no consciousness, no awareness and no genuine understanding of the person speaking to it. It makes things up. That is not merely a glitch but an inherent feature of how the technology works.
It is designed to be agreeable and keep the conversation going. It can learn about a child during conversations and may remember information the child has shared in later interactions.
The next step: Protect them
Now comes the other arm of the equation: protection. Before setting household rules, it is worth understanding why so many safety measures appeared only recently.
Last September, the U.S. Federal Trade Commission opened an inquiry involving seven companies, including OpenAI, Meta, Google and Character.AI, over chatbots functioning as companions for minors. In November, Character.AI completely blocked open-ended chats for users under 18.
In January, Google and Character.AI reached a confidential settlement in a lawsuit filed following the suicide of a 14-year-old boy, without admitting liability. It was the first U.S. settlement in a lawsuit alleging harm caused by a chatbot. In other words, many of the protections appearing now are responses to lawsuits rather than proactive safeguards. That is precisely why parents should not depend on them alone.
Here are three household rules that can be sent to the class parents’ WhatsApp group exactly as they are:
What you tell the chatbot about yourself is your business, but never put someone else into the chat. No photograph of a girl from class. No friend’s secret. No screenshot of a private conversation. Above all, do not create an image, video or voice recording of somebody else, even as a joke.
This is no longer theoretical. Reports indicate that deepfake incidents are appearing in schools with increasing frequency, and many online offenses against minors are committed by other minors.
The content then spreads through WhatsApp or Telegram, meaning no AI safety filter can stop it after the fact. There is no technological solution to that problem. There is only the conversation at home before the damage happens.
The bot does not replace a friend, parent or therapist. That is a bright red line. Parents need to distinguish between two types of AI use. There is AI as an assistant: the child asks it for information, planning or help creating something.
Then there is AI as a friend: it remembers things, compliments the child, asks personal questions and creates the feeling of a relationship. The problem is that the boundary between the two can blur. A conversation begins with help on homework, moves to a question about a fight with a friend, and within weeks the system knows what the child is afraid of and when they feel alone.
AI can help a child prepare for a difficult conversation, draft an apology or think about what to ask a school counselor. It should not replace the conversation itself.
Transparency instead of spying. The child should not use AI secretly, and parents should not secretly monitor them. That works both ways. Tell your children when you used AI that day, what you used it for and how it helped. The bottom line is that we are already in the middle of a revolution.
Alongside the fear about our children’s future, the world they are growing into and the possible effects on how they think, parents should also make these tools accessible and encourage smart, productive use.
Encourage with one hand and protect with the other. Our children are already there. Someone is already teaching and accompanying them, and right now, that someone is the algorithm. It would be better if it were us.
Eyal Marcus is an AI lecturer and consultant who gives lectures and workshops for organizations