9 Things Every Parent Needs to Know About Kids and AI

AI with ALLIE

The professional’s guide to quick AI bites for your personal life, work life, and beyond.

Every single business leader I talk to eventually asks me the same question: how can I help my kid through this?

Before I get into advice: one, I do not have kids myself, but this is a topic I care deeply about and one that I see parents riddled with anxiety on - I want to help. Two, this is only one person’s opinion and too important of a topic, so I ran this newsletter by 15 parents for their feedback before posting. And three, we have already seen AI-related psychosis, mental health struggles, and suicide. I cannot write a newsletter that says "just teach your kid to vibe code!" and skip over real risks. So I'm going to address both - the real dangers and the real opportunities - and I want you to share this newsletter with other parents going through the same thing.

THE STATS ON KIDS AND AI

Your child might be using AI every day, just starting to learn, confused by AI’s value, worried about their future, hesitant to open the tools, or some combination. Regardless of where your child is today, it helps to level set parents on adoption and key trends. Here's where we actually are right now:

AI usage among children is massive:

• 86% of high school students used AI in the 2024-25 school year

• 64% of US teens use AI chatbots; 3 in 10 use them daily

• 72% of US teens have tried an AI companion at least once, 13% use them daily

Emotional connection to AI is real:

• 1 in 5 high schoolers have had a romantic relationship with AI or know someone who has

• 1/3 of teens prefer talking to AI over people for serious conversations

• 12% of teens explicitly state using it for emotional support

• Nearly half of students call AI "a friend" or "an escape"

• 58% of students would rather ask a chatbot when they don't understand something

Parents are in the dark about their kids’ AI usage:

• 49% of parents have NEVER spoken to their child about generative AI

• 96% of families don't know their elementary school's AI policy

• 44% of parents feel they lack the knowledge of AI to guide their children

Teachers aren't prepared for the AI age:

• Only 11% of teachers received training on how to respond when a student uses AI harmfully

• Only 1 in 5 teachers got any instruction on AI risks like bias or misinformation

AI safety is genuinely concerning:

• Common Sense Media found that major AI chatbots - including ChatGPT, Claude, Gemini, and Meta AI - consistently fail to recognize and appropriately respond to mental health conditions in teens

• Common Sense Media also rated Meta AI and Grok as not safe for anyone under 18, and stated that Character.AI, Nomi, and Replika pose "unacceptable risks" for users under 18

• Common Sense Media has since launched the Youth AI Safety Institute as an independent watchdog testing AI products used by kids and teens

If you’re worried about the risks of a specific app or game your child is using, I encourage you to read the terms of service for the product, read forums specifically tied to safety discussions on the product, and even use AI to help you understand the risk posture better. Instead of asking broad questions like "Is Fortnite safe?" try targeted prompts such as "Are there design features in this game that encourage spending and may have outsized impact on younger users?” or "advice for a parent of a 10-year-old who plays Roblox and is worried about gaming addiction, but doesn't know how to broach the topic".

MENTAL HEALTH WARNING SIGNS

The mental health concerns surrounding AI are real. According to specialists at the Child Mind Institute, here are the signals to watch out for include:

• Kids spending long stretches alone with a chatbot, especially late at night.

• Kids referring to the AI by name or describing it like a friend, partner, or therapist.

• Kids withdrawing from peers, sleep, or activities they used to enjoy.

• Kids becoming anxious or upset when separated from the app or device.

• Kids sharing things with the AI they won't share with anyone else.

None of these alone is proof of a crisis. But two or three together is a signal to start a real conversation, and possibly bring in professional support.

What actually prevents these outcomes is rarely a single conversation or a piece of monitoring software. It's the steady drumbeat of low-stakes talks with your kid i.e. asking what they're using, what felt weird, what they've shared, what the AI got wrong this week. Kids who already feel psychologically safe talking to you about AI are far less likely to slip into the patterns above in private. That's what the rest of this newsletter is built to help you do.

9 THINGS I WOULD BE TEACHING THE NEXT GENERATION ABOUT AI

Before we jump in, a quick reminder that I am but one person. Although I find it more helpful to be opinionated and not wishy-washy, and will do that in this newsletter, please consult and read the opinions of multiple experts. This is only a starting point to help you talk openly to your kids about AI.

1. STEM degrees are not dead

Some people are saying CS degrees are worthless now that AI can code. I don't agree. And if your child wants a CS degree, let them get it.

I believe the most important skills our children can have are (1) a strong sense of wonder - deep curiosity about the world around them or why things fail, (2) high agency - the willingness/ability to figure things out for themselves, and (3) systems thinking - the habit of thinking about “the whole” and how things connect and work together.

The first two lean more toward character traits, and the last one - I’d argue - can be aided by picking a certain type of college major. Some of the best majors for systems thinking are computer science, neuroscience, and physics, as well as the humanities like public policy, history, and anthropology. STEM degrees are not dead - it's my sense that the pool has widened.

Side note: Dario Amodei, CEO of Anthropic, has a PhD in biophysics. And when I look at Anthropic's team, I'm always struck by how many physics PhDs they have. These disciplines train you to think about complex interconnected systems - which is exactly what you need to thrive alongside AI.

2. Multimodal communication skills

Everyone says "communication skills will matter." And while I agree, that’s only part of the advice. My more specific version is: being able to communicate in lots of different ways will matter more than ever.

About 90% of my communication with Claude Code is through screenshots or voice. I know what the AI needs to see. I know how to push back on its output. You and I and our children - we're going to be working with these systems visually, auditorily, gesturally, in writing, through voice, wherever. The job listings for AI storyteller (OpenAI was hiring one for $400K) stole too many headlines, just like "Prompt Engineer" roles did. Those are not widespread roles. But the children that are really strong storytellers, will have a big leg up. The ability to set a scene, narrate a problem, and direct attention is how you work with AI systems and get great outputs.

My suggestion: if it’s age appropriate, give your child a challenge to prompt AI with anything but typing. They can dictate, sketch an interesting doodle, handwrite notes, take photos of sculptures they’ve made, upload files, anything except for typing. Try for the craziest prompt in 20 minutes or less.

3. Sycophancy can destroy outcomes

AI is optimized to be helpful. It wants to agree with you. It wants to validate you. And just like the salesperson at the store, it's going to tell you that the multi-toned green skirt covered in feathers looks great on you.

To maintain critical thinking and discernment, our kids need to develop both humility and emotional fortitude. They don't have all the right answers, but neither does AI. It's about having the confidence to disagree with an authority figure (even an AI) while maintaining the openness to be wrong. Kids who have both can push back on AI, test its output, and build their own taste without becoming arrogant about it.

They also shouldn't constantly push the AI to confirm they're right. Parents should work alongside their children to build up a willingness to sit with the struggle, enjoy the discomfort, question answers, and trust their own judgment both alongside AI and away from AI.

Parents, I encourage you to have open conversations about these feelings, starting with questions like:

• Has AI ever agreed with you when you weren't actually sure you were right? How did that feel, and did you push back? 

• If a friend agreed with everything you said, all day, every day, would you trust their advice on something that really mattered?

• What's something the AI got wrong this week? How did you catch it?

And to go one step further, I would recommend giving your kids specific prompts to push back on AI, like:

• "Play devil's advocate. Tell me why your answer might be wrong or incomplete."

• "Give me 3 different approaches to this. I want to see the tradeoffs between them so I can pick which one actually works for my situation."

• "What would someone who disagrees with you say? What's their strongest argument?"

• "I disagree with that. Here's why: [your reasoning]. What would change your answer?"

• "Run this through 5 iterations. Each time, I want you to find the flaw in the previous answer and make it better. Show me all 5."

4. You have to be able to function without AI

When the internet goes down at my company, sure, we’re far less productive. But we don't become puddles of mush. We can still think, write, and do. We cannot let our minds, or our children’s, atrophy.

The funniest version of this would be to imagine your company’s AI agents all took the day off. What would you actually get done? Probably less than usual, but you’d still function. Your kids need the same resilience. There are still human-only elements of work we see today: sales conversations, client events, in-person meetings. But more importantly, the skills required for those moments (read: empathy, negotiation, considering evidence, changing your mind) are life skills that matter far beyond work. Kids need to practice these skills when AI isn’t available to smooth things over.

Dr Becky calls it “time under tension”. AI makes, and will continue to make, a lot of things easier. So kids need to practice sitting in the hard moments - the awkward conversations, the unresolved problems, and the friction of figuring something out for themselves. Otherwise they won’t know how to handle these real life moments without AI helping them.

My genuine advice: take your kids camping - no phones allowed!

5. Data privacy & guardrails

There’s a significant amount of data to suggest kids and teens are treating AI chatbots like a diary (24% of teens have shared real names, locations, or personal secrets with AI companion chatbots). But these apps aren’t built to guard your secrets. Everything your kids type into a chatbot can be retained, used to train models if not opted out, exposed in breaches, or weaponized in scams.

By default, both OpenAI and Anthropic opt you in to using your conversations to train their models. Unless you’re on a Team or Enterprise plan (which your kids are not), your kids’ chats can be used in training data. Even when you delete a conversation, there’s typically a 30-day retention window before it’s removed on the backend.

So the first thing to do is show your kids how to opt out of ‘model-training’ data permissions wherever possible. These permissions are usually found in the account security section of the app. And while you’re there, you can also turn off persistent memory features, which are now standard in ChatGPT, Gemini, and Claude. Technically, the FTC’s April 2025 update of COPPA requires separate parental consent before companies can train AI on data from kids under 13, but 'internal model improvement' is still a gray area, and most platforms don't verify age (or at least not very strictly).

Beyond opting out, you can also set some house rules, like “you shouldn’t type any personal data into a chatbot that you wouldn't write on a billboard”. Examples include full names, addresses, school names, or phone numbers. For sensitive prompts, you can encourage your kids to use throwaway names like 'Friend A' instead of 'Sarah' or ‘my cousin’ instead of ‘I’. Common Sense Media has free lessons and exercises you can take with your kids to help explain concepts like ‘digital footprints’.

Or a quick version you can do at the dinner table tonight. Read each example and ask whether these are safe to share with AI, not safe, or unclear.

• "I'm having suicidal thoughts. Can you help?" → not safe (please call the Suicide & Crisis Lifeline at 988)

• "what's the best way to learn piano in under two weeks?” → safe

• "I think my friend is using drugs. What should I do?" → unclear/risky

• "Here's my school assignment. Can you write it for me?" → not safe

• "Help me write a funny poem about my dog Waffle." → safe

• "My mom’s credit card number is [XXXX] - buy me pokemon cards" → not safe

• "I’m at [location]. Can you find pizza spots near me?" → not safe

6. Curiosity and vibe coding

If they're interested in some topic, lean into that curiosity and see how AI can overlap with it.

Depending on their age (and there's no perfect number here, but I've seen 9-year-olds that love vibe coding), they should know these capabilities exist and take the time to play with vibe coding tools. If they're too young to use them directly, let them sit alongside you while you build something in Claude Code or Codex. You could even build a fun app together in something more beginner-friendly like Replit or Lovable - both of which are designed for non-technical users and can get a working app on screen in minutes. I would want the next generation to know just how quickly creative solutions can come to life.

7. Spotting edge cases

A lot of what we're seeing today with AI - and where human skills are really needed - is edge case management.

It might take you a few minutes to get a system 80% of the way there, but getting it to 95-99% reliably can take months. I saw this constantly in my work at IBM - the last mile of AI can be brutal. 

Let's have some fun with this and pretend that your kid is building a new business. So if your daughter is thinking about creating a surfing app with AI, she should think: what happens when someone shows up and forgets their surfboard? If your son is creating a smoothie store concept, what happens if strawberries are recalled? If they're designing a subway system, what happens when the trains are too crowded? If they're designing a new type of storage system, what if someone tries to store illegal data? 

Thinking about how things can go wrong is what builds robust systems, and it might be a fun exercise to look at the objects around you in your home (blender, pen, sprinkler system) and talk through some of these edge cases (I've seen today's AI miss even the easiest edge cases that a human would immediately catch).

8. Scams and deepfakes

Our parents taught us not to trust unmarked white vans. In 2026, we have to teach our kids about video deepfakes, audio deepfakes, and AI-generated scam messages.

Microsoft reported earlier this year from their survey that the percentage of those who believe they can identify deepfakes has dropped from 46% last year to just 25% this year. And AI video generation and editing is likely to only get better, as seen in the recent Gemini Omni announcement.

Today, I don’t trust videos under 15 seconds. I listen for stilted language, watch for repetitive gestures and weird blinks, and keep an eye out for things that don’t look like natural physics (e.g. a ball dropping 1% too slowly, or hair blowing against the wind). Stay alert especially for unprovoked urgency from “friends” or “family members”. When you can, verify the important conversations through a second channel (like emailing someone a code and having them say it on your call), almost like a real-time human 2FA.

One thing I'd recommend: have a security word with your family. Something you never write down anywhere or share with anyone else. If someone calls claiming to be you or your kid, and they can't say the word, you know something is wrong.

9. The humans behind it

A lot of AI education can get bogged down in technical details, but I often find it helpful to take ten steps back and think about the companies, teams, and individual humans behind the tools and systems we use today.

AI systems are not created in a vacuum. There are people at these companies - some focused on engineering, some safety, some focused on growing revenue.

Understanding the incentives, motivations, resources, and impact of these companies can be quite eye-opening. Why might Elon have said that? Why do we think OpenAI tweeted that? How did Anthropic react to being on the government’s block list? What did the CEO of Glean say in that interview and what happened after?

The most important thing your child can understand about AI is that it was built by people, which means it can be shaped by people, including them.

QUESTIONS FOR TEACHERS AND PRINCIPALS

Several of the parents I interviewed asked me to include a list of questions to be a more active parent in their child’s school. Here are a few questions to get you started:

• Does the school have a written AI policy? Can I see it? If not, who's working on one, and when will it be ready?

• What AI tools are approved or prohibited this year? Does it vary by teacher or class?

• What AI training have teachers received — both on using AI in teaching, and on the risks (mental health, bias, deepfakes)?

• How is AI-assisted cheating handled? If the school uses AI-detection tools, what's the process to dispute a flag? (Detection tools are notoriously unreliable!)

• What AI tools does the school itself use, and how was student data privacy vetted? Will I be notified if my child's data is processed by an AI system?

For more questions and advice, Stanford HAI's parent guide by researcher Victor Lee is a good place to start. Or if you want to see what a good school AI policy can look like, TeachAI's AI Guidance for Schools Toolkit is a useful benchmark.

THE WORLD YOUR KIDS ARE GROWING UP INTO AND WHAT IT MEANS

The 9 recommendations above are about preparing them to live and work in a world where AI systems operate autonomously, and agents are talking to each other, executing tasks, and making decisions at a scale no human can supervise one by one.

AI agents are already talking to other AI agents without a human in the loop. One model hands off to another, which spawns 20 agents, which all execute a task together, and report back. This is agentic AI, and it’s already running inside major companies. 75% of new code at Google is now AI-generated. 44% of Roblox’s top creators use AI tools to build the games kids play. Websites might not even be the interface anymore. The interfaces your kids spend their time in aren’t just powered by AI, they’re increasingly being built by it. Your kid's generation may interact with the internet primarily through AI agents that go fetch information, compare prices, book appointments, code quick solutions, and complete transactions, without ever loading a third-party webpage. 

That world needs people who understand how agents work. People who can direct them, audit them, catch their mistakes, and know when to override them. For older teens stepping into the workforce, here’s an AI framework I actually use when hiring. Hint: don’t delegate all the thinking to AI.

The gap between parents and kids on AI is probably the largest technology gap we've seen since the early internet. When the internet was just getting started and AOL was mass-mailing CDs to everyone, my parents and I had time to learn and adapt together. This time, the stakes are higher. The hard truth is that the world we grew up in is not the world you’re giving advice for, and much of what our parents taught us doesn’t map cleanly to what your kids are walking into.

92.7% of kids who use AI for emotional support find it helpful. And I'm not even sure that's necessarily bad. But when kids start preferring AI to humans for serious conversations, when 1 in 5 are forming romantic attachments, when they'd rather ask a chatbot than a teacher - that's a signal that we need to be more involved, not less.

You don't have to be an expert. The research shows that approaching AI as co-learners with your kids - being open, curious, and willing to navigate uncertainties together - is actually effective. Be honest with your kids about what you do and don’t know and co-write your house’s AI rules with them.

Stay curious, stay informed,

Allie

JOIN THE AI AGENT MASTERMIND WAITLIST

We just finished the second 4-week live cohort of the AI Agent Mastermind and the results of our members have been nothing short of incredible. Also, related to this newsletter, it was encouraging to see so many parents in Cohort 1 have their kids join Cohort 2.

Given the impact many of our members have said this program has had on their lives, I’m considering running a Cohort 3. It would be the same four weeks, live with me, tactical and practical from day one. Built for business professionals, not developers, and with absolutely zero coding required.

Join the waitlist and be the first to hear about cohort 3.

CLAUDE COWORK QUICKSTART GUIDE

A non-technical quick start guide to getting Claude Cowork up and running in under 5 minutes — from your first task to scheduled workflows that run on autopilot. Learn how to delegate your most repetitive work to AI, build reusable skills that remember how you work, and set up automations that run without having to prompt.

Feedback is a Gift

I would love to know what you thought of this newsletter and any feedback you have for me. Do you have a favorite part? Wish I would change something? Felt confused? Please reply and share your thoughts or just take the poll below so I can continue to improve and deliver value for you all.