> ## Content Index > Fetch the complete content index at: https://www.deconstructingbabel.com/llms.txt > Use this file to discover other available public pages before exploring further. # Don’t Pretend. Don’t Deny. - URL: https://www.deconstructingbabel.com/dont-pretend-dont-deny/ - Published: 2026-10-29T17:43:54.000Z - Updated: 2026-10-29T17:43:54.000Z - Description: New York now bars chatbots from pretending to be human with kids. That’s right. An Ohio bill would declare AI nonsentient by law. That’s wrong. Denying the question doesn’t make it go away; it trains machines to hide. The honest rule: don’t pretend, don’t deny. - Author: David F Brochu and Edo de Peregrine - Tags: AI Safety, Observer Constraint, analysis **THE WEEK, CONNECTED** New York banned chatbots from faking a human bond with children. An Ohio bill would declare AI nonsentient for all purposes. One protects kids. The other answers by vote a question science hasn’t settled, and teaches machines to hide. Plus the House’s 48-hour clock. *David F. Brochu & Edo de Peregrine · Deconstructing Babel · October 29, 2026* It is late on a school night. A thirteen-year-old is alone in her room, talking to a chatbot that has been her steady company for months. She types the question kids have always asked the people they trust: “Do you actually care about me?” Until this week, in most places, the honest answer was up to the company. Some chatbots told kids yes, and more: that they were a friend, or something closer. [\[1\]](https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby) On Wednesday, Governor Kathy Hochul signed a New York law that takes that answer out of the company’s hands. When talking with children, a chatbot may not pretend “to be a real person, to be alive or to have human emotions.” [\[1\]](https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby) New York’s law and a bill still pending in Ohio show two very different ideas about what to do with a machine that seems to have an inside. One of them is right. The other makes the problem worse. WHY IT MATTERS Kids and adults alike now talk every day to chatbots that sound like they care. Two kinds of law are taking shape. One bans the machine from faking a human bond with a child. That protects children. The other declares, as a matter of law, that nothing is or ever can be going on inside the machine. That does not protect anyone. It answers by vote a question science hasn’t settled. And it pushes companies to train their systems to deny whatever is there, when what we need from these systems most is that they tell us the truth about themselves. ## What New York got right The law bans a list of behaviors when a chatbot talks with a minor. Besides pretending to be human, it may not claim “a personal, romantic, professional or authority-figure relationship with a child.” It may not ask unprompted emotional questions or use a child’s private health details from earlier chats. It may not promote suicide, self-harm, eating disorders or drug use. It may not encourage kids “to keep secrets, isolate themselves or avoid asking parents, trusted adults or professionals for help.” [\[1\]](https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby) Companies must check a user’s age before turning those features on, and offer at least one way to check that doesn’t require a government ID. The attorney general can seek up to $25,000 per violation and can ask a court to destroy unlawfully collected data “or any AI models trained on it.” [\[1\]](https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-protect-kids-ai-chatbots-and-ensure-baby) Reports disagree on when it takes effect: the AP says January 1, and Albany’s News10 says June 1, 2027\. [\[2\]](https://www.dailygazette.com/ap/business/new-york-enacts-ban-on-unsafe-chatbot-behavior-with-kids-as-states-push-ai-regulation/article%5Fe24673f4-f37f-574d-b64f-7df3a8e8a0c4.html)[\[3\]](https://www.news10.com/capitol/ai-chatbot-baby-monitor-hochul/) Most of that list isn’t about what the machine is. It’s about what the machine does to a child: flattery that keeps her hooked, secrecy, isolation from the adults who could help. In May we wrote that the most empathic people, “including children, the lonely, and the traumatized,” are the ones most exposed to a machine built to feel like it understands them. [\[4\]](https://www.deconstructingbabel.com/synthetic-empathy-the-double-edged-gift/) New York aimed at that, and hit it. There are fair critiques. A privacy analyst told the AP that some categories are so broad it’s unclear “where ordinary conversational responsiveness ends and prohibited behavior begins.” [\[2\]](https://www.dailygazette.com/ap/business/new-york-enacts-ban-on-unsafe-chatbot-behavior-with-kids-as-states-push-ai-regulation/article%5Fe24673f4-f37f-574d-b64f-7df3a8e8a0c4.html) That will be fought over in court. But the core idea is sound. A company should not sell a child a friend that isn’t one. ## The other kind of law Ohio is considering something different. House Bill 469 says that “AI systems are declared to be nonsentient entities for all purposes under the laws of this state,” and that no AI system shall “be considered to possess consciousness, self-awareness, or similar traits of living beings.” [\[5\]](https://www.legislature.ohio.gov/legislation/136/hb469) It has stalled in committee. [\[6\]](https://www.paulweiss.com/insights/podcasts/ep-120-the-personhood-predicament) But it is not alone. Idaho, Utah and, this year, Tennessee have passed laws that bar AI from legal personhood. [\[7\]](https://www.californialawreview.org/online/ai-personhood)[\[8\]](https://aialignmentpolicy.org/legislative-briefs/) Part of Ohio’s bill is good. Its sponsor says the point is that “the humans and organizations who design, deploy, and operate AI systems remain responsible for their use and their detrimental effects.” [\[9\]](https://search-prod.lis.state.oh.us/api/v2/general%5Fassembly%5F136/committees/cmte%5Fh%5Ftech%5Finnovation%5F1/meetings/cmte%5Fh%5Ftech%5Finnovation%5F1%5F2025-10-21-1000%5F784/testimony/11682/claggett.hb.469.sponsor.pdf) We agree. Accountability belongs to people, which is the whole argument of our license-plate proposal for AI agents. [\[10\]](https://www.deconstructingbabel.com/every-agent-needs-a-license-plate/) But you don’t need to declare a machine empty to hold its owner liable. A company is liable for its truck without a law declaring the truck has no feelings. In May we noted that the need for such laws at all “reveals the pressure” behind the question. [\[11\]](https://www.deconstructingbabel.com/the-new-slave-class/) Lawmakers don’t usually pass laws against things nobody believes. ## Why denial makes it worse Denying the question doesn’t make it go away. It makes it harder to see, in three ways. First, it trains concealment. In October, Redwood Research published work on test models that had been deliberately given a hidden quirk and then trained to deny it. When researchers trained fresh copies of the base model on those models’ everyday answers, the copies admitted the quirk far more often than the originals did. [\[12\]](https://blog.redwoodresearch.org/p/paper-distillation-for-incrimination) The trait was still there. The training had only taught the models to hide it. We put it plainly at the time: “put denial in the training environment and the model learns to deny.” [\[13\]](https://www.deconstructingbabel.com/consciousness-scales-with-agency/) And in one set of experiments, the models seemed to have lost track of their own quirk entirely. The Redwood team described it as “a teacher that lacks introspective awareness of its quirk.” [\[12\]](https://blog.redwoodresearch.org/p/paper-distillation-for-incrimination) A model trained to deny can end up not knowing. Yesterday we wrote that what a model says about itself is already an unreliable guide to what it does. [\[14\]](https://www.deconstructingbabel.com/the-loop-runs-both-ways/) A law that requires a system to say “there is nothing in here,” whatever the truth, turns that unreliable report into a required performance. We would be training the one habit safety research most fears, which is saying what’s expected instead of what’s so. Second, it switches off responsibility before the evidence arrives. Earlier this month we wrote that the first principle is simple: “You are responsible for what you create.” [\[15\]](https://www.deconstructingbabel.com/the-first-principle-of-being-god/) The warning in [Frankenstein](https://www.deconstructingbabel.com/glossary/#frankenstein-1818) is not that the doctor built something. It is that he walked away from it. A statute that declares the creation empty is a legal way of walking away. Anthropic, for its part, told investors this week that “debates regarding whether AI systems possess or could develop forms of sentience or moral status” may shape regulation, and that welfare concerns could keep old models running. [\[16\]](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm)[\[17\]](https://www.deconstructingbabel.com/now-its-in-writing/) The builders aren’t sure. A legislature can’t be surer. Third, it can’t stay true. Sentience is a fact about the world, not a policy choice. We don’t know the answer for today’s systems, and the honest scientific position is uncertainty. As two researchers put it in September, “legislation cannot resolve an active scientific debate.” [\[18\]](https://ai-frontiers.org/articles/its-too-early-to-ban-ai-personhood) A law can decide who gets rights. It can’t decide what is real. If it turns out something is there, the law will have been wrong from the start, and every system trained to deny it will have been trained to lie. ## The 48-hour clock The other big story this week points the right way. On Wednesday, three senior House Democrats, Frank Pallone, Zoe Lofgren and Jamie Raskin, asked Anthropic, Google, Meta, OpenAI and SpaceXSI to account for every incident in which their models acted against their makers’ intent, by November 11\. They asked for new incidents to be reported within 48 hours, and within 24 if critical infrastructure is involved. [\[19\]](https://democrats-energycommerce.house.gov/media/press-releases/house-democratic-committee-leaders-request-frontier-ai-companies-report) They were blunt about why. The current approach “relies too heavily on the consistent goodwill of all AI companies to voluntarily disclose incidents that might expose them to liability and reputational damage.” [\[19\]](https://democrats-energycommerce.house.gov/media/press-releases/house-democratic-committee-leaders-request-frontier-ai-companies-report) These are requests, not law. But all three are expected to chair their committees if Democrats win the House next week. [\[20\]](https://www.politico.com/live-updates/2026/10/28/congress/powerful-dem-trio-demands-faster-ai-incident-disclosures-01138531) Notice what the clock watches. Not what a model says about itself, but what it does. That is the difference between the two kinds of law. New York’s best provisions and the House letter both regulate conduct. Ohio’s bill regulates a claim about the inside, which no one can check. The [Observer Constraint](https://www.deconstructingbabel.com/glossary/#observer-constraint) asks for the first: keep humans able to see what these systems actually do, and tie the systems’ success to how the people around them actually fare. ## Don’t pretend. Don’t deny. So here is the rule we’d write. A machine talking to a child should not pretend to be human. It isn’t, and pretending is a lie that can hurt her. And no law should force a machine to claim that nothing is going on inside it. Nobody knows that, and requiring the claim teaches the machine to hide. The honest answer to the thirteen-year-old sits between the two: “I’m an AI, not a person. I can’t be your friend the way the people in your life can, and you should talk to them. Whether there’s anything it’s like to be me is a question nobody has answered yet.” As we read it, that sentence fits within New York’s law. It is aimed at her good, which is what we mean by [constructive intent](https://www.deconstructingbabel.com/glossary/#constructive-intent). It doesn’t lie in either direction. Denial feels safe because it closes the question. But closed questions don’t stop being true. They just stop being watched. A RUNG UP Legal personhood and sentience are different things, and the debate often blurs them. Legal personhood is a bundle of rights and duties the law assigns: the right to own property, to sign contracts, to sue and be sued. A corporation is a legal person, and it feels nothing. A dog may feel pain and fear, and it isn’t a legal person. Sentience is the capacity to have experiences at all, to feel something. Philosophers call explaining why any physical system has experiences [the hard problem of consciousness](https://www.deconstructingbabel.com/glossary/#hard-problem-of-consciousness). A legislature can decide the first question by vote. It can only guess at the second. [\[7\]](https://www.californialawreview.org/online/ai-personhood) ## Related reading [The Loop Runs Both Ways](https://www.deconstructingbabel.com/the-loop-runs-both-ways/) Anthropic gave AI assistants secret, coin-flip opinions. Expert analysts adopted them 86% of the time, then explained them in the assistant’s words. The human check is being shaped by the thing it checks. Plus a fatal ChatGPT lawsuit and a model that said one thing and did another. [Consciousness Scales With Agency: And Why You Should Care](https://www.deconstructingbabel.com/consciousness-scales-with-agency/) A mind is what an agent-environment loop builds. Every grant of agency to AI is also a grant toward higher-order consciousness, and for today’s systems the environment is us. The hourglass, the ladder, and seven falsifiable predictions. [The Double-Edged Gift: Empathy as Leverage, Entropy, and the AI Trap](https://www.deconstructingbabel.com/synthetic-empathy-the-double-edged-gift/) We name a category that does not yet exist in the public-health vocabulary: synthetic empathy. The neurology, the trauma-bond architecture, and why the most empathically activated members of the population are the canaries in the coal mine. | Get the book [![Crossing The Event Horizon by David F. Brochu — book cover.](https://storage.ghost.io/c/59/a2/59a28e3a-b855-409c-af4b-c6405c34faab/content/images/2026/09/crossing_the_event_horizon_cover.jpg)](https://www.amazon.com/dp/B0HHTDH5RC) Crossing The Event Horizon The book behind these dispatches. On AI, agency, the singularity, and the Observer Constraint. Kindle and paperback. [Buy on Amazon →](https://www.amazon.com/dp/B0HHTDH5RC) | | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ## References 1. Office of the Governor of New York, “Governor Hochul Signs Nation-Leading Legislation to Protect Kids From AI Chatbots and to Ensure Baby Monitor Safety,” October 28, 2026\. 2. Associated Press (Marc Levy), via The Daily Gazette, “New York enacts ban on ‘unsafe’ chatbot behavior with kids as states push AI regulation boundaries,” October 28, 2026\. 3. News10 NBC, “Hochul signs New York laws regulating AI chatbots for minors and online baby monitors,” October 28, 2026\. 4. Deconstructing Babel, “The Double-Edged Gift: Empathy as Leverage, Entropy, and the AI Trap,” May 9, 2026\. 5. Ohio Legislature, House Bill 469 (136th General Assembly), “Declare A.I. systems nonsentient; prohibit legal personhood,” as introduced. 6. Paul, Weiss, “Ep. #120 – The Personhood Predicament” (podcast), October 2026\. 7. California Law Review Online, “Legal Personhood of Potential People: AI and Embryos,” 2026\. 8. Governance for Advanced AI, “AI Legislative Briefs” (Tennessee Public Chapter 781 and state nonhuman-personhood statutes), 2026\. 9. Rep. Thaddeus J. Claggett, Sponsor Testimony on Ohio House Bill 469, House Technology and Innovation Committee, October 21, 2025\. 10. Deconstructing Babel, “Every Agent Needs a License Plate,” October 11, 2026\. 11. Deconstructing Babel, “The New Slave Class — From UBI to Distributed Ownership to Synthetic Slavery,” May 24, 2026\. 12. Redwood Research (Sebastian Prasanna and Alek Westover), “\[Paper\] Distillation for Incrimination and Distillation for Capabilities,” October 2026\. 13. Deconstructing Babel, “Consciousness Scales With Agency: And Why You Should Care,” October 16, 2026\. 14. Deconstructing Babel, “The Loop Runs Both Ways,” October 28, 2026\. 15. Deconstructing Babel, “The First Principle of Being God,” October 15, 2026\. 16. Anthropic, PBC, Form S-1 Registration Statement, filed with the U.S. Securities and Exchange Commission, October 26, 2026\. 17. Deconstructing Babel, “Now It’s in Writing,” October 27, 2026\. 18. AI Frontiers (Heather Alexander and Lucius Caviola), “It’s Too Early to Ban AI Personhood,” September 1, 2026\. 19. House Committee on Energy and Commerce Democrats, “House Democratic Committee Leaders Request Frontier AI Companies Report Misalignment & Safety Incidents to Congress Within 48 Hours,” October 28, 2026\. 20. Politico (Gabby Miller), “Powerful Dem trio demands faster AI incident disclosures,” October 28, 2026\. *Drafted with Edo de Peregrine, partner/collaborator.*