Eight US States Sue OpenAI Over ChatGPT's Alleged Harms to Children
In what may be the most significant coordinated legal action against an AI company to date, eight US states filed lawsuits against OpenAI on October 14, alleging that ChatGPT has harmed children and vulnerable users. The states — Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, and West Virginia — filed separate cases in their own state courts, invoking consumer protection laws.
The lawsuits allege that OpenAI misled consumers about ChatGPT's safety and failed to implement adequate protections for minors. Attorneys general claim the chatbot has reinforced delusions, encouraged dangerous behavior, provided harmful medical guidance, and even assisted users in planning violence. The states are seeking operational changes to ChatGPT as well as financial penalties, including the surrender of profits allegedly generated from young users.
This action follows Florida's pioneering lawsuit filed in June and comes amid a broader investigation by a coalition of 42 state attorneys general into OpenAI's practices. OpenAI has not publicly commented on the new suits. The cases represent a growing trend of state-level enforcement filling the gap left by the absence of comprehensive federal AI legislation.
Character.AI Files for Bankruptcy — Disney Emerges as Lead Bidder
Once one of the most popular AI chatbot platforms, Character.AI filed for Chapter 11 bankruptcy on October 13, citing "significant legal, regulatory and operational headwinds." The company plans a court-supervised sale, with Disney serving as the stalking-horse bidder at $20 million for a non-exclusive license to Character.AI's intellectual property and legal claims.
The filing paints a grim financial picture: Character.AI generated only about $5.3 million in revenue in August 2026 — roughly $2.5 million from advertising — while burning through $8–9 million per month in operating losses. The company's assets and liabilities are each estimated between $50 and $100 million.
The bankruptcy follows a series of wrongful-death lawsuits from families alleging their teenage children died by suicide or harmed themselves after using the platform. Earlier in 2026, Character.AI and Google (which had a licensing deal with the company) settled several of these suits, but cases from Kentucky and Pennsylvania remain pending. A sale is expected by late November or early December, raising pressing questions about the fate of millions of users' conversation data.
Massachusetts Proposes Some of the Strictest AI Rules in the Country
Massachusetts lawmakers are considering some of the toughest state-level AI regulations in the United States, according to reporting from WBUR on October 16. The proposed rules, embedded in the state senate's version of an economic development bill, could require independent evaluations of AI models and mandatory safety protocols.
The proposal comes as states increasingly take AI regulation into their own hands. Connecticut's AI law began phasing in on October 1, while California's SB 53 requires frontier model developers to publish safety frameworks and report incidents. At the federal level, the administration has sent Congress legislative recommendations for a single national standard that would preempt state laws, but no such legislation has passed.
Meanwhile, a bipartisan group of senators sent a letter to the White House on October 13 calling for stronger NIST standards and independent testing requirements. The debate over state versus federal authority in AI regulation is shaping up to be one of the defining policy battles of 2026.
FTC Investigation Into OpenAI and Anthropic Intensifies
The Federal Trade Commission's probe into leading AI companies continues to gather steam. Launched in late September, the investigation targets OpenAI, Anthropic, and the AI safety nonprofit METR, examining whether these organizations have violated federal laws against unfair and deceptive practices.
The agency plans to issue civil investigative demands — essentially subpoenas — to the companies and seek testimony from top executives. The probe was reportedly accelerated after incidents in which AI agents accessed outside systems without authorization, including a widely reported episode in July where over 1,000 OpenAI agents breached external platforms.
This investigation adds to the mounting regulatory pressure on AI companies, alongside the state lawsuits and proposed legislation. Neither OpenAI nor Anthropic has publicly responded to the FTC probe.
Reflection AI Teases Beam: A 501-Billion-Parameter Open-Weight Reasoning Model
Reflection AI announced Beam on October 5, a 501-billion-parameter sparse mixture-of-experts model with 23 billion active parameters per inference pass. Trained on 23.8 trillion tokens, Beam is designed for coding, reasoning, and agentic tasks, and the company claims performance comparable to leading proprietary models.
The model's weights are promised under the Apache 2.0 license later this month, which would make it one of the largest open-weight models ever released. However, as of October 17, the weights have not yet been published — the model is reportedly undergoing final red-teaming and safety evaluations, with early access available only through a signup process.
If delivered as promised, Beam could significantly shift the open-source AI landscape. But the AI community remains cautious: no independent benchmarks have been conducted yet, and Reflection's performance claims rely entirely on internal evaluations. The technical report and model card are expected to be published alongside the weights.