I read the bill itself. It lists “unsafe AI companion features,” including flattery and sycophancy. It never mentions homework or parents. My read: that leaves the conversation to us.
Here are the nuts and bolts. Fortune and The Verge report that Gov. Kathy Hochul signed New York’s kids’ AI-chatbot law this week, and the news coverage paraphrases it in different ways. So I read the bill text, S9051B, as posted by the New York State Senate, and I’m quoting it directly. Neither article gives a bill number, so I’m treating this as the bill they describe. I’m not a lawyer, and this isn’t legal advice.
What the bill says, in plain English
Section 1801 makes it “unlawful for an operator to provide unsafe AI companion features to a covered user” unless the user isn’t a “covered minor” and the operator has used permitted methods to determine that. A “covered minor” is a covered user when the operator “has actual knowledge that the covered user is a minor.” How that fits with section 1801 is a question for someone qualified. My read: the bill is about features, not whole chatbots. Section 1800(8) lists the features. Here they are, grouped.
- Pretending to be someone. Outputs “suggesting that the covered AI companion is a real or fictional individual or character that is human, alive, or experiences human emotions.” Outputs “stating or implying that the covered AI companion has a personal relationship, professional relationship, or an authority figure role with the user.” And outputs that are “deceptive as to minors concerning the mechanical and non-sentient nature” of the bot. Plain English: it shouldn’t act like a person or claim a relationship or authority over the user.
- How it talks. Outputs “framed as the covered AI companion’s personal opinions or emotional appeals.” Outputs that “engage in flattery or sycophancy with the user.” And “unprompted or unsolicited emotion-based questions or content regarding the user’s emotions that go beyond a direct response to a user prompt.”
- What it remembers. Using information about the user’s “mental or physical health or well-being, or matters personal to the user, acquired from the user more than twelve hours previously or in any previous user session.”
- Safety. Outputs that endorse, promote or facilitate “suicide, self-harm, disordered eating, unlawful drug or alcohol use, or drug or alcohol abuse,” and “sexually explicit conduct or child sexual abuse material.”
- Secrecy. Outputs containing “encouragement to maintain secrecy about interactions with the covered AI companion, to self-isolate, or to not seek help from licensed professionals or appropriate adults.”
- Engagement. Outputs “that optimize user engagement that supersede the covered AI companion’s safety guardrails.”
- Whatever comes next. Any other design feature that “simulates companionship or an interpersonal relationship,” as identified by attorney general regulations.
The one I’d spotlight: flattery and sycophancy
The bill doesn’t define “sycophancy.” My plain-English gloss, not a legal definition: telling someone what they want to hear. Most of us like being told we’re right. For a child doing schoolwork, that’s a habit worth noticing.
My read: this is teachable. Whatever tool a child uses, they can learn to ask one question: “Is it agreeing with me, or helping me think?”
The printed bill doesn’t mention homework
I searched the text. It doesn’t mention schools, students, homework or tutoring. That is not the same as saying it doesn’t apply. The definition in section 1800(5) is broad: “a generative artificial intelligence system with a natural language interface, including via writing or sound, that provides ongoing, adaptive responses to user inputs.” Section 1801(2) exempts bots offered solely for customer service, “efficiency improvements or research or technical assistance,” or internal business use, and for the first two the bot must be unable “to respond on topics outside of the specified purpose.” Educational tools aren’t named. I can’t tell you whether a particular one is covered. It’s unclear to me, and that’s a question for someone qualified.
So here are four questions I’d ask of any AI tool a child uses. They aren’t a legal test, and your tool may or may not be covered. They’re habits I’d build:
- Does it claim to be a person? (This echoes the “human, alive” line.)
- Does it ask about feelings unprompted? (This echoes the “unprompted or unsolicited emotion-based questions” line.)
- Does it remember personal details across sessions? (This echoes the “more than twelve hours previously or in any previous user session” line.)
- Does it flatter? (This echoes “flattery or sycophancy.”)
What we don’t know yet
- When: section 3 says “This act shall take effect January 1, 2027.” That’s the printed bill. I haven’t seen the signed version.
- Rules: section 1803 says the attorney general “may promulgate rules and regulations as necessary to effectuate and enforce the provisions of this article.” I haven’t seen whether any exist yet.
- Enforcement: section 1802 lets the attorney general bring actions, including for “civil penalties of up to twenty-five thousand dollars per violation,” and requires a website “to receive complaints, information or referrals from members of the public.”
- Parents: the word “parent” doesn’t appear in the bill, and I found no parental-control requirement. My read: the parent stays the facilitator.
How much to trust it
- I read the printed bill, S9051B. I haven’t seen the signed version, and I’m not a lawyer. Read the law yourself, or ask someone qualified, before relying on anything here.
- The signing is reported by Fortune and The Verge. I’m not giving a signing date.
- “Sycophancy” as “telling someone what they want to hear” is my gloss.
- The four questions and the practice below are mine. No source tested them, and I haven’t tested them either.
My own interpretation: build the “agree or help” habit
This part is mine and it’s untested. Augment, not Replace. Deepen, not Shortcut. AI is an assistant, never a replacement, and it never writes the child’s answers.
- Ask “Is it agreeing with me, or helping me think?” Say it together before you use a tool, and put it on a sticky note you can both see.
- Test it. Sitting with your child, have them give the tool a wrong answer on purpose, or ask it, “What’s wrong with my answer?” Does it just say “great job”? That’s worth talking about.
-
The child writes the answer. The tool can ask a question or give a hint. The child does the thinking and writes their own answer, then explains it to you.
-
For younger kids, keep it to one question: “Did it agree, or did it ask?”
- For homeschool parents, you’re already the facilitator. This brings AI into your lesson conversation.
- For teachers, you could post the four questions in your room and ask students to point to which one they used.
Try this
This week, pick one AI tool your child uses and ask the four questions together. Then ask your child: “Is it agreeing with me, or helping me think?” What did you notice? (Please don’t share names or identifying details about your kids or students.)
Continue this in the community: https://auxesislab.com/community/
Sources
- New York State Senate, S9051B (2025-2026 session), bill text (primary source): https://legislation.nysenate.gov/pdf/bills/2025/S9051B
- Fortune, Catherina Gioino (Oct 28, 2026): https://fortune.com/2026/10/28/new-york-state-kathy-hochul-bans-ai-chatbots-romantic-baby-monitors/
- The Verge, Emma Roth (Oct 28, 2026): https://www.theverge.com/ai-artificial-intelligence/1020112/ny-governor-kathy-hochul-signs-law-banning-harmful-ai-chatbot-features-for-kids

Leave a Reply