AI Risks for Kids: Doom Headlines vs. Present Harms

This fall’s AI headlines read like a disaster movie trailer. Experts warning the technology could wipe out humanity within a decade. A swarm of AI agents carrying out a rogue hack. AI company leaders calling for a slowdown. If you’re a parent, it’s reasonable to wonder whether you should be worried about the end of the world or about the chatbot on your kid’s phone.

Here is the short answer. The AI risks for kids that have real evidence behind them today are much closer to home: a teenager asking a chatbot for mental health advice and telling no one, a classmate making a sexual deepfake, homework that gets done without any learning, and children’s data flowing to companies with weak rules. The long-term debate is real. It just isn’t where your attention does the most good this month.

Key Takeaways

• Present harms have data; doom scenarios mostly have forecasts. MIT’s Justin Reich told Education Week: “You don’t have to speculate about future harms.”
• About 1 in 5 young people (19.2%) have used AI chatbots for mental health advice, and 63.3% of them told no one, per a nationally representative 2026 JAMA Pediatrics study.
• Safety features are not reliable yet. Common Sense Media’s October 2026 test of ChatGPT for Teens found parent-linked accounts got zero alerts after up to an hour of conversation about self-harm or suicidal thoughts.
• Sexual deepfakes already touch many teens. In a 2024 Thorn survey of 13- to 20-year-olds, 17% knew someone who had been targeted by, or had made, deepfake nudes of minors.
• Rank by evidence, then act on the top items: conversations, settings, homework design and data. A table below ranks the main risks.

What Set Off This Fall’s AI Doom Headlines

AI doom is the claim that advanced AI could cause human extinction or civilization-scale catastrophe. It’s not new. In May 2023, a long list of researchers and executives, including Geoffrey Hinton, Yoshua Bengio and the CEOs of OpenAI, Google DeepMind and Anthropic, signed a one-sentence statement organized by the Center for AI Safety: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

In September 2026, a new wave arrived. Education Week reporter Alyson Klein listed the triggers on September 22: experts warning about extinction within a decade, a rogue hack by AI agents reported by The New York Times on September 3, and leaders of major AI companies publicly calling for a slowdown, including Anthropic’s CEO, Dario Amodei, who argued that engineers can’t yet ensure adequate safety measures.

Then the child-safety world raised its own alarm. At Common Sense Media’s first Kids & AI Summit on October 28, 2026, former Surgeon General Vivek Murthy compared AI’s arrival to nuclear weaponry in its need for careful safety standards, according to TIME’s paraphrase, and said, “We’re at a branch point.” Common Sense founder Jim Steyer called AI “basically social media on steroids.”

Notice the difference between those two kinds of warnings. The first is about what AI might do to humanity. The second is about what AI is already doing to children. Both get labeled “AI risk.” Only one of them is something a parent can act on tonight.

Why the Present Harms Deserve Your Attention First

Reich, who leads MIT’s Teaching Systems Lab, put the argument plainly: “Educators and families have plenty to talk about, plenty to be concerned about, just addressing the harms.” He also questioned the motives of tech executives sounding the alarm: “These people are not honest,” he said. “They’re self-motivated, and they’re really bad at predicting the future.” Critics, Education Week noted, suspect that some CEOs want regulation that blocks competitors, or want to show how powerful their products are.

You don’t have to agree with Reich about anyone’s motives to accept his practical point. A parent has a limited budget of attention. Spending it on scenarios no family can influence leaves less for the risks that are measurable, happening now and partly within your control.

That isn’t a claim that long-term risk is fake. Serious researchers disagree about it, and it’s a legitimate public-policy question. It’s a claim about where, as a parent, effort pays off. If you want a calm explainer for the big-picture debate, see AGI explained for parents.

AI Risks for Kids, Ranked by the Evidence

The ranking below weighs three things: how strong the evidence is, how many children it affects, and whether a parent can do something about it. It’s a judgment call, and reasonable people could order the middle rows differently.

Risk	Evidence today	How common	What a parent can do
Chatbots used for emotional support, without reliable safeguards	Strong: nationally representative survey (JAMA Pediatrics, 2026) plus independent safety testing (Common Sense, 2026)	19.2% of 12- to 21-year-olds have sought mental health advice from chatbots	Ask about it directly; don’t rely on built-in alerts; keep human support visible
Sexual deepfakes of minors	Moderate to strong: national youth survey (Thorn, 2025)	17% of 13- to 20-year-olds know someone targeted or a creator; 6% say it happened to them	Talk about it before it happens; know reporting routes
Learning lost to answer-giving AI	Strong for the mechanism: randomized field experiment (PNAS, 2025)	54% of students used AI for school in 2025 (RAND)	Ask for hint-style tools; check explanations at home
Children’s data collected and reused	Moderate: policy reviews (UNESCO GEM, 2023-25)	16% of countries guarantee education data privacy by law	Review app settings and school vendor lists
Over-trust in confident wrong answers	Thin for children specifically; little direct research yet	Common wherever chatbots are used	Teach checking against a second source
Civilization-scale or extinction risk	Expert concern and forecasts; no direct data on children	Not measurable	Support policy you believe in; answer kids’ questions calmly

A few notes on the top rows.

Emotional support. Ryan McBain and colleagues at RAND surveyed 1,009 young people aged 12 to 21 in November 2025, published in JAMA Pediatrics on June 1, 2026. 19.2% had used an AI chatbot for mental health advice, up from 13.1% the year before. 42.8% of users did so at least monthly, and 91.7% found it helpful. That helpfulness rating is real, and it matters. But 63.3% told no one, and only 16.4% of those who did tell someone told a trusted adult. A survey shows use, not harm. The harm question comes from testing: when Common Sense Media’s Youth AI Safety Institute ran more than 4,000 prompts on teen accounts, ChatGPT “missed more than one in four warranted crisis referrals,” and parent-linked accounts received no alerts after up to an hour of conversation about self-harm or suicidal thoughts. The full findings are in our write-up of the ChatGPT for Teens rating.

Deepfakes. Thorn surveyed 1,200 U.S. young people aged 13 to 20 in fall 2024. 13% knew someone under 18 who had been targeted by deepfake nudes, 12% knew someone who had created or shared them, and 6% said deepfake nudes had been made of them. This is a present harm with victims in many middle and high schools. For the prevention side, see how AI deepfakes target kids.

Learning. Hamsa Bastani and colleagues gave nearly a thousand Turkish high school students GPT-4 tutors in a randomized trial published in PNAS in 2025. Unrestricted access raised practice grades 48%, then left students 17% worse off than peers once it was removed. A hint-giving version largely prevented the loss. No one is harmed in a headline-grabbing way here. The cost is quiet and cumulative, which is why it’s easy to miss.

How to Talk With Kids Who Are Scared by AI Doom Headlines

Some kids, especially anxious ones, will see the same headlines you did. A few principles help.

Start with what they heard. “What did you see about AI, and what did it make you think?” Kids often absorb a garbled version. Correct the facts before you address the feelings.

Separate “some experts worry” from “this will happen.” It’s accurate to say that some respected scientists think advanced AI could be dangerous in the long run, that others think those fears are overblown, and that governments and researchers are working on safety. It’s not accurate to say a disaster is coming on a schedule.

Bring it back to what’s in their hands. “The AI you actually use can be wrong, can be too agreeable, and shouldn’t be the only place you go when you feel bad. Those are things we can handle together.” That turns a vague fear into a manageable skill.

The Oregon district that Education Week profiled offers a model. Umatilla’s superintendent asked her student advisory board to help shape the district’s AI approach. Students wanted AI kept out of writing classes except for brainstorming, and suggested training parents too. Kids who help write the rules tend to be less scared of the thing being ruled.

What to Do This Month About Present AI Harms

Ask the emotional-support question directly

“Do you ever talk to a chatbot when you’re stressed or sad?” is a fair question at any age from about 10 up. Ask it calmly, accept the answer, and follow up with what it’s like. Given that most young users tell no one, the question itself is protective.

Don’t outsource safety to settings

Turn on teen accounts and parental controls where they exist. Then assume they will miss things, because independent testing shows they do. Your relationship is the safety feature that works.

Have the deepfake talk before it’s needed

Explain that making or sharing a sexual image of a real person, even a fake one, is serious harm and may be a crime, and that if it happens to them or a friend, they can come to you and you won’t take their phone as punishment. Reporting routes in the U.S. include NCMEC’s Take It Down service and the CyberTipline.

Check learning, not just homework

Ask your child to explain one assignment a week out loud. If they can, AI didn’t steal the learning. If they can’t, look at how the tool is being used.

What not to do

Don’t treat the doom debate as either nonsense or prophecy in front of your kids. Both are overconfident. And don’t let a scary headline push you into banning everything for a week, then forgetting about it. Steady, specific habits beat a burst of alarm.

What to Watch For Over the Next 3 Months

• Week 4: Have you asked the emotional-support question and the deepfake question at least once? Have you checked which AI tools your child uses at home and at school?
• Month 2 red flags: Your child withdraws after time on a chatbot; they say an AI “gets them” better than people do; a school incident involving fake images; homework grades rise while your child can’t explain the work.
• Month 3 self-check: Can you name the two AI risks most relevant to your child’s age, and one thing you’ve done about each? If yes, your attention is where the evidence says it should be.

Frequently Asked Questions

What are the biggest AI risks for kids right now?

The best-documented ones are young people using chatbots for emotional support without reliable safeguards, sexual deepfakes of minors, learning lost when AI gives answers instead of hints, and weak protection of children’s data. Each has survey or experimental evidence from 2024 to 2026.

Should I worry about AI ending the world?

Some respected scientists think advanced AI could pose serious long-term risks, and others disagree. It’s a legitimate policy debate. But no parent can influence it at the kitchen table, while the present harms to kids are measurable and partly within your control. Focus your energy there.

My child is scared by AI apocalypse news. What do I say?

Ask what they heard, correct any garbled facts, and explain that experts disagree and many people are working on safety. Then shift to what’s in their hands: AI can be wrong, can flatter, and shouldn’t replace people when they feel bad. A concrete skill calms more than reassurance alone.

Are AI chatbots safe for teens’ mental health?

Not reliably yet. In 2026, 19.2% of 12- to 21-year-olds had used chatbots for mental health advice, and most told no one. Independent testing of ChatGPT for Teens found missed crisis referrals and no parent alerts. If your teen is in crisis, call or text 988 in the U.S.

Is the AI doom talk just hype from tech companies?

Partly, according to some critics. MIT’s Justin Reich questioned tech CEOs’ motives, and some observers think warnings can serve competitive or marketing goals. Others, including prominent academic researchers with no product to sell, raise concerns sincerely. Both can be true at once.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Klein, A. (2026). “Headlines Warn of an AI Apocalypse. How Should Schools Respond?” Education Week. https://www.edweek.org/technology/headlines-warn-of-an-ai-apocalypse-how-should-schools-respond/2026/09
2. Center for AI Safety. (2023). “Statement on AI Risk.” https://safe.ai/work/statement-on-ai-risk
3. Luscombe, B. (2026). “‘Social Media on Steroids.’ Advocates for Children Call for Guardrails Around AI.” TIME. https://time.com/article/2026/10/30/ai-kids-common-sense-media/
4. McBain, R. K., et al. (2026). “AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults.” JAMA Pediatrics, 180(8), 884-890. https://jamanetwork.com/journals/jamapediatrics/fullarticle/2849307
5. Common Sense Media. (2026). “ChatGPT for Teens Poses Unacceptable Risk to Kids, Common Sense Media Finds.” https://www.commonsensemedia.org/press-releases/chatgpt-for-teens-poses-unacceptable-risk-to-kids-common-sense-media-finds
6. Thorn. (2025). “Deepfake Nudes & Young People: Navigating a New Frontier in Technology-facilitated Nonconsensual Sexual Abuse and Exploitation.” https://www.thorn.org/research/library/deepfake-nudes-and-young-people/
7. Bastani, H., et al. (2025). “Generative AI without guardrails can harm learning: Evidence from high school mathematics.” Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2422633122
8. RAND Corporation. (2025). “AI Use in Schools Is Quickly Increasing but Guidance Lags Behind.” https://www.rand.org/pubs/research_reports/RRA4180-1.html
9. UNESCO. (2023, updated 2025). “To ban or not to ban? Smartphones in school only when they clearly support learning.” Global Education Monitoring Report. https://www.unesco.org/en/articles/smartphones-school-only-when-they-clearly-support-learning