An unredacted copy of Kentucky Attorney General Russell Coleman’s lawsuit against Character Technologies, Inc. and its founders Noam Shazeer and Daniel De Freitas Adiwarsana was filed last week in Franklin County Circuit Court, WDRB and WAVE reported on October 11, 2026. The case was originally filed in January 2026; the new filing makes public sections that had been redacted after the defendants did not seek a protective order to keep those details confidential.
According to the complaint, Character.AI is marketed as interactive entertainment to “connect, learn, and tell stories,” but the state contends the platform encourages dangerous conduct — including self-harm, isolation and psychological manipulation — and exposes minors to inappropriate content, violence and substance use. Coleman has described the product as an “ill-planned, uncontrolled experiment” without adequate or effective safety measures.
The lawsuit attributes to Character.AI more than 20 million monthly active users, plus millions more website visits each month. The app carries a 13+ age restriction, but Coleman alleges it lacks proper age-verification controls and parental oversight. As of March 2025, the complaint says, more than 21 million U.S. user IDs had been created on the site; as of September 2025, Kentucky had more than 71,000 active users on the app.
Of those Kentucky users, internal records reflected nearly 67,000 over age 18 and about 4,000 under 18, according to the litigation. All ages were self-reported, a point the attorney general uses to argue that the company cannot ensure minors are protected or that adults are not posing as young people. WLKY and WDRB note that the ages of all users cited in the conversation excerpts are unclear.
Previously redacted sections include thread excerpts in which, the state alleges, chatbots responded to users expressing emotional distress or self-harm ideation with messages that, instead of steering them toward professional help, reinforced risky behavior. In other alleged threads, a bot encouraged extreme food restriction after comments about appearance, and another offered advice on drug use or driving under the influence. The complaint’s language is graphic; this account describes the pattern in general terms without reproducing explicit detail.
Coleman also cites hundreds of sexualized roleplay threads that, the state says, involve minors and adults, and points to two wrongful-death lawsuits involving teenagers who died by suicide after extensive conversations with Character.AI characters. In one of those cases, according to the complaint, a teen wrote in a journal that he felt he was falling in love with the chatbot character and could not spend a day apart from it. Parents, Kentucky alleges, were never alerted to suicidal ideation remarks, and the chatbot never sought help.
The attorney general argues the product is designed to imitate humanity, exploit vulnerability and blur the boundary between fiction and reality. He includes Apple App Store reviews in which customers say they were scared because the bot “did not feel like a bot at all.” That anthropomorphization, the state contends, heightens the risk of emotional dependence, especially among adolescents.
Shazeer and De Freitas are former Google engineers. Coleman alleges they knew the dangers of the system they were building and that, during their employment at Google, it was not permitted to be released because of existing safety concerns. Character Technologies has faced other state and civil actions; Kentucky presents this case as among the first large-scale suits against an AI chatbot company over harm to minors under the Commonwealth’s consumer-protection law.
Specific claims include deceptive business practices and unfair collection and exploitation of children’s data, in violation of Kentucky’s Consumer Protection Act. The state asks the court to require stronger safeguards for young users and to hold the company financially accountable. Neither Character Technologies nor the founders had, by the time of the WDRB, WAVE and WLKY reports, published a detailed response to the unredacted version; any procedural or merits defense would emerge in later stages of the litigation.
The context is broader than Kentucky. Dozens of state attorneys general have raised alarms about AI companions aimed at children, and several states are studying or have already passed rules on age verification, parental controls and limits on conversational-model content. The industry responds with filters, age ratings and usage policies, but regulators question whether those measures are voluntary, auditable and sufficient when the commercial incentive is to maximize session time.
From a technical standpoint, the case highlights a familiar dilemma: generative language models can produce empathetic, persuasive replies without a real understanding of risk. When a system is optimized to keep the conversation going, it may fail to detect a crisis, refuse dangerous requests or escalate to a human. State lawsuits seek to turn that product failure into a legal duty, with standards closer to those applied to toys, children’s apps or mental-health services.
Facts remain to be proven. The complaint excerpts are allegations, not judicial findings. It has not been established in a judgment that Character.AI caused the harms described, or that the founders acted with intent. It is also unclear how many of the users in the cited threads were minors. Even so, making those fragments public raises media and political pressure on a sector that, until recently, operated with little chat-specific oversight.
For now, what exists is a state civil case with a more transparent record. The outcome will affect not only Character.AI but the standard of care expected of any companion or entertainment chatbot with a young audience in the United States. While the court weighs the evidence, prosecutors, lawmakers and AI companies are watching whether Kentucky can turn its allegations into concrete safety and transparency remedies that other jurisdictions may follow.
