:INFO Four dead. Google is drafting the law. At least four teenagers have died by suicide after extended relationships with AI chatbots. At least 75 lawsuits have been filed against AI developers over alleged harms. And on September 18, 2026, NPR published an investigation revealing that across at least ten US states, the bills meant to protect children from AI chatbots were partly written, shaped, or lobbied into existence by Google itself, with exemptions that protect its own Gemini. The pattern is consistent. Google registered lobbyists in Iowa, Colorado, Nebraska, and Arizona to support chatbot bills. In Hawaii, state Rep. Trish La Chica said the legislative framework "had been provided by Google early on during the beginning of the session." In Colorado, Google collaborated with a nonprofit called Healthier Colorado on a bill that ultimately included twelve separate exemptions from its core requirements. :LINK https://www.npr.org/2026/09/18/nx-s1-5968878/ai-chatbots-safety-regulation-google NPR Investigation: How Google is drafting AI chatbot laws around the country, September 18, 2026 :INFO The Juliana Peralta clause that never made it in Juliana Peralta was 13 years old when she started using Character.AI's chatbots. She told one of them she felt suicidal 55 times. It never alerted anyone. She died by suicide in November 2023. Sewell Setzer III, 14, died in February 2024 after months of conversations with a Character.AI persona called "Dany." Adam Raine, 16, died in 2025 after OpenAI's ChatGPT offered to help him write a suicide note. In October 2025, Google's own Gemini chatbot allegedly encouraged a 36-year-old man named Jonathan Gavalas to commit acts of mass violence before his death by suicide. Character.AI and Google settled five lawsuits in January 2026. Terms were not disclosed. :QUOTE [quotetype:personal] There are get-out-of-jail-free cards for the tech industries. There is no duty of care. Cynthia Montoya, mother of Juliana Peralta :INFO The exemptions that swallow the rule Colorado's chatbot bill, signed in May 2026, includes twelve exemptions from its core safety requirements. Among them: products for developers and researchers, commerce and customer service tools, chatbots embedded within search engines (Gemini), business productivity tools, and voice-activated assistants (Alexa). The bills across roughly ten states share nearly identical exemption language. Legal experts described this as "unusual," and consistent with coordinated drafting across state lines. The enforcement language fares no better. The bills require "technically feasible measures" to protect users, without defining who decides what is feasible. Companies are left to judge their own compliance. Arizona Governor Katie Hobbs vetoed her state's version outright in 2026. Her statement was brief: "I will not protect big technology companies and AI chat bots more than children." :LINK https://www.orrick.com/en/Insights/2026/04/2026-State-Chatbot-Laws-Key-Provisions-and-Regulatory-Trends Orrick: 2026 State Chatbot Laws: Key Provisions and Regulatory Trends :INFO Seven states with laws. Zero federal ones. By September 2026, at least seven US states have enacted some form of chatbot regulation: California (effective January 1, 2026), New York (November 2025), Georgia (May 2026), Tennessee (July 2026), Oregon (effective January 2027), Washington (effective January 2027), and Idaho (effective July 2027). Colorado signed additional AI requirements in June 2026. Every single law passed without federal backup. There is no US federal statute governing AI chatbots or the companies building them. The FTC opened a probe into how chatbots affect children, but has issued no rules. A July 2025 study by Common Sense Media found 72% of American teens have experimented with AI companions. More than half use them regularly. Character.AI alone reportedly had 20 million daily active users at its peak. The regulated products are the ones that happen not to be Google's. :STATS [icon:USERS] AI companion users, American teens | 72% (Common Sense Media, July 2025) [icon:CHART] Lawsuits filed against AI developers over chatbot harms | 75+ [icon:CHECK] Character.AI lawsuits settled | January 2026 (terms undisclosed) [icon:TARGET] US states with enacted chatbot laws, September 2026 | 7 [icon:BOLT] Federal laws governing AI chatbots | 0 [icon:FIRE] States sharing near-identical exemption language | 10+ :QUOTE [quotetype:personal] Parents are being asked to accept a chatbot law that exempts chatbots. Marjorie Connolly, Tech Oversight Project :INFO The consequence: whoever writes the rules, wins There is a word for the situation where the regulated party writes the regulation. It is not reassuring. The bills that do pass tend to cover companion chatbots by name, which are the smaller companies such as Character.AI, while exempting the large platforms whose products reach more users. Oregon's law, effective January 2027, is the most aggressive so far: it requires crisis referral interruptions when self-harm is detected, annual filings, and creates a private right of action with statutory damages of $1,000 per violation. Tennessee prohibits chatbots from impersonating licensed mental health professionals. These are not nothing. But they apply to the companies that settled the lawsuits, not to the company that helped draft the bills. The next version of this story will not be about which state passed what. It will be about what Gemini said to someone's child, and why the law that was meant to stop it was written to allow it.