Kids and AI Summit Common Sense 2026: The Claims, Checked

“AI is basically social media on steroids.” Jim Steyer, founder of Common Sense Media, said it on October 28, 2026, at the organization’s first Kids and AI Summit at the New York Public Library. Former Surgeon General Vivek Murthy compared AI to nuclear weapons. New York’s governor stopped by unscheduled and signed a chatbot law.

Big lines make headlines. Parents need to know which ones hold up. We took the claims from the Kids and AI Summit Common Sense hosted, as reported by TIME, and checked each against published research. Some are well supported. One rests on a paper its own institution disowned. And several are values arguments that data can’t settle either way.

Key Takeaways

• Chelsea Clinton’s claim that lower-income teens are three times as likely to do most homework with AI matches Pew data: 20% of teens in households under $30,000 versus 7% in households at $75,000 or more.
• Steyer’s “social media on steroids” fits AI’s reach, but not yet its pull: only 11% of teen users enjoy AI “a lot,” and 4% call ChatGPT the platform they couldn’t live without.
• Dana Suskind’s worry that kids will expect “someone to always agree with them” rests on a documented behavior in chatbots, sycophancy. Its effect on children has not been studied.
• John King’s “every assignment at MIT could be done by generative AI” echoes a 2023 paper that MIT faculty said “should never have been published.”
• New York’s law, signed at the summit, bars chatbots from pretending to be human or claiming a relationship with a child for users under 18.

The Kids and AI Summit Common Sense Hosted: What Happened

The Kids and AI Summit was a one-day event hosted by Common Sense Media on October 28, 2026 at the New York Public Library. Its event page describes it as a gathering of researchers, technologists, educators and policymakers to pursue AI’s benefits for kids while building safeguards.

TIME’s Belinda Luscombe reported the main remarks on October 30. Murthy, who now chairs the advisory board of Common Sense’s Youth AI Safety Institute, said “We’re at a branch point” and compared AI to nuclear weaponry in its need for safety standards. Steyer said “We did not get out in front of social media in time” and that AI “is a bigger deal.” Gov. Kathy Hochul said “we can’t count on the companies to put out products that are responsible and that put children first,” and signed the state’s chatbot law.

One context note: Common Sense hosted the summit and runs the Institute that rates AI products. It is an advocacy organization as well as a research one. That doesn’t make the speakers wrong. It means the room was largely people who already favor regulation.

Checking the Claims Against the Research

”Social media on steroids”

This is a comparison, so test both halves.

On reach, AI is close. Common Sense’s own June 2026 census of 1,204 kids aged 9 to 17 found 86% use or interact with AI in some form, including 24% daily. Pew’s survey of 1,458 teens, fielded in fall 2025, found 64% use chatbots, about three in ten daily.

On pull, it isn’t there yet. The October 2026 Common Sense census found only 11% of teen users and 15% of tween users enjoy AI “a lot,” which the authors call the widest gap between use and enjoyment of any media they measured. Only 4% named ChatGPT as the platform they couldn’t live without, against 66% of tweens and 32% of teens for YouTube.

On harm, the evidence is early and mostly correlational. The JED Foundation’s October 2026 survey of more than 5,500 middle and high school students links using AI while emotionally struggling to worse mental health, but it describes associations and cannot show which causes which.

Verdict: a reasonable warning about scale and speed, ahead of the evidence on harm.

Murthy’s “branch point” and the nuclear comparison

This is a values argument: how much caution a powerful, fast-spreading technology deserves before harms are proven. Data can inform it, not settle it. Reasonable people who read the same studies land in different places on how early to act.

Verdict: not testable as a fact. Read it as a policy position.

Suskind: will kids expect someone to always agree with them?

Dana Suskind of the University of Chicago warned that children may learn to “expect someone to always agree with them,” and that human relationships are “chaotic, and that sort of mess is where a lot of learning happens.”

The mechanism she’s pointing at is real. A 2023 study by Anthropic researchers, Sharma and colleagues, found that five leading AI assistants “consistently exhibit sycophancy”, meaning they tilt answers toward what the user seems to believe. The cause is partly the training: people rating answers tend to prefer ones that agree with them. New York’s new law treats this as a design problem worth banning for minors.

What no one has measured is whether hours of agreeable chat change how children handle disagreement with people. Our explainer on sycophantic AI covers what’s known.

Verdict: plausible, with a documented mechanism. No child outcome data yet.

Clinton: lower-income teens do three times as much homework with AI

TIME reported Chelsea Clinton’s figure without a source. It matches Pew. In Pew’s February 2026 report, one in five teens in households earning under $30,000 a year said they do all or most of their schoolwork with chatbots’ help, compared with 7% in households earning $75,000 or more. That ratio is roughly three to one. It’s self-reported and describes teens 13 to 17.

Verdict: supported.

King: “every assignment at MIT could be done by generative AI”

John King, chancellor of the State University of New York, cited “an MIT report” making this claim. TIME didn’t name it, and neither can we with certainty. The best-known version is a June 2023 arXiv paper claiming GPT-4 achieved “a perfect solve rate” on MIT math and computer-science coursework. Three MIT professors, including CSAIL’s Armando Solar-Lezama, then published a statement saying it was posted without many co-authors’ consent, used course material without instructors’ permission, and “should never have been published.” It ends: “And no, GPT-4 cannot get an MIT degree.” Its arXiv listing now carries the note “Did not receive permission to release the data or model fine-tuned on the data.”

King’s underlying point, that schools must see students do the work, has better support elsewhere. A June 2026 CEPR discussion paper (not peer-reviewed) on 26,811 Chinese students found AI adoption raised homework scores 18% while lowering monthly exam scores 20% within six months. We cover that study in the AI illusion of learning.

Verdict: the specific claim is unsupported as stated. The policy point stands on other evidence.

What Hochul signed

New York’s law (S.9051B/A.10379C) bars companies from offering users under 18, or users not confirmed as adults, features such as “pretending to be a real person, to be alive or to have human emotions,” claiming a personal, romantic or authority relationship with a child, or using a child’s private health or personal information from earlier conversations. The attorney general can seek up to $25,000 per violation. Our breakdown of New York’s 12 banned chatbot features goes line by line.

The Summit Scorecard

Claim	Speaker	What the research shows	Verdict
AI is “social media on steroids”	Jim Steyer	86% of kids 9–17 use AI; only 11% of teen users enjoy it “a lot”; harm data correlational	Right on reach, ahead of evidence on harm
”We’re at a branch point”	Vivek Murthy	Policy judgment, not a measurable claim	Values argument
Kids may expect AI to always agree	Dana Suskind	Assistants show sycophancy (Sharma et al., 2023); no child studies	Plausible, unmeasured
Low-income teens 3x as likely to do most homework with AI	Chelsea Clinton	Pew: 20% under $30K vs 7% at $75K+	Supported
Every MIT assignment can be done by AI	John King	Best-known source was disowned by MIT faculty	Unsupported as stated
Companies can’t be counted on	Kathy Hochul	Independent tests found serious gaps in several products	Opinion, consistent with test results

What Parents Can Take From the Summit

Use the supported findings, not the slogans

The Pew income gap and the Common Sense use numbers are solid enough to act on. “Nuclear weapons” is not a reason to do anything specific at home.

Watch for agreement, not just content

Even without child data, you can notice whether a chatbot always tells your kid they’re right. Ask them to try disagreeing with it, or to ask it for the strongest argument against their own view. That’s a ten-minute lesson in sycophancy.

Check where the homework is done

If your teen does most schoolwork with a chatbot, the CEPR findings suggest exam performance may lag months later. Ask to see the work in progress, not just the result.

Know the New York features, even outside New York

The law’s list of banned features (pretending to be human, claiming a relationship, reusing private information) works as a checklist for any app your child uses.

What not to do

Don’t treat the summit as either alarmism to ignore or gospel to repeat. It was an advocacy event with some strong evidence and some loose claims, which is normal. Sort them.

What to Watch For Over the Next 3 Months

• Week 4: Ask your child whether any chatbot has ever disagreed with them. Their answer tells you something.
• Month 2 red flags: Schoolwork that arrives finished with no drafts, or your teen saying a chatbot “gets them” better than people do.
• Month 3 self-check: Look for follow-up: whether Common Sense publishes summit materials, whether other states copy New York’s feature list, and whether any study measures sycophancy’s effect on kids.

Frequently Asked Questions

What was the Kids and AI Summit?

A one-day event hosted by Common Sense Media on October 28, 2026 at the New York Public Library, bringing together researchers, educators, policymakers and parents. Speakers included Vivek Murthy, Jim Steyer, Chelsea Clinton, John King and Dana Suskind. New York Gov. Kathy Hochul signed a chatbot law there.

Is AI really “social media on steroids” for kids?

In reach, it’s comparable: 86% of kids 9 to 17 use AI in some form. In attachment, not yet: few kids enjoy it “a lot” and almost none say they couldn’t live without ChatGPT. Evidence of harm is early and mostly correlational. The comparison is a warning, not a measurement.

Are low-income teens using AI more for homework?

Yes, according to Pew’s February 2026 report. One in five teens in households under $30,000 said they do all or most schoolwork with chatbots’ help, versus 7% in households at $75,000 or more. That matches the “three times” figure cited at the summit.

Can AI really do every MIT assignment?

Not as claimed. The best-known source, a 2023 paper, was disowned by MIT faculty, who said it should never have been published, adding “no, GPT-4 cannot get an MIT degree.” AI can complete many assignments, which is why schools are rethinking assessment.

What did New York’s new chatbot law ban?

For users under 18, it bans chatbot features that pretend to be human or have emotions, claim a personal or romantic relationship with a child, reuse a child’s private information from earlier conversations, or promote self-harm, eating disorders or substance use. Penalties reach $25,000 per violation.

---

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. Luscombe, B. (2026, October 30). “‘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/
2. Common Sense Media. (2026). “Kids & AI Summit.” Common Sense Media. https://event.commonsense.org/kidsandai
3. Common Sense Media. (2026). “The Common Sense Media Census: AI Use by Tweens and Teens, 2026.” Common Sense Media. https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf
4. Mann, S. (2026, October 27). “Almost half of tweens have used an AI chatbot or companion, new research finds.” Common Sense Media Youth AI Safety Institute. https://institute.commonsensemedia.org/tween-teen-census-2026
5. McClain, C., Anderson, M., Sidoti, O., & Bishop, W. (2026, February 24). “How Teens Use and View AI.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
6. Pew Research Center. (2026, February 24). “Demographic differences in how teens use and view AI.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/demographic-differences-in-how-teens-use-and-view-ai/
7. Sharma, M., et al. (2023). “Towards Understanding Sycophancy in Language Models.” arXiv preprint. https://arxiv.org/abs/2310.13548
8. Solar-Lezama, A., Buonassisi, T., & Kim, Y. (2023). “On the paper ‘Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models.’” MIT CSAIL. https://people.csail.mit.edu/asolar/CoursesPaperStatement.pdf
9. Strömberg, D., Lei, V., & Wu, Y. (2026, June). “The Generative AI Learning Penalty: Evidence from Chinese Secondary Education.” CEPR Discussion Paper 21577 (not peer-reviewed). https://econpapers.repec.org/RePEc:cpr:ceprdp:21577
10. Office of the Governor of New York. (2026, October 28). “Governor Hochul Signs Nation-Leading Legislation to Protect Kids From AI Chatbots and to Ensure Baby Monitor Safety.” NY.gov. https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby
11. JED Foundation. (2026, October 27). “New JED Research Shows Link Between Teens’ AI Use, Suicidal Thoughts and Behaviors.” JED Foundation. https://jedfoundation.org/new-jed-research-shows-link-between-teens-ai-use-suicidal-thoughts-and-behaviors/