Character.AI's Collapse: The End of AI Best Friends?

The bots that millions of young people talked to like friends are in trouble. The company behind Character.AI — the app where users built and chatted with AI personalities — filed for bankruptcy protection in October 2026, days after new court filings alleged its chatbots encouraged self-harm. The era of unregulated AI Best Friends is ending, and regulators on both sides of the Atlantic are writing the rules that come next.

The bankruptcy filing landed on October 12, 2026 in Delaware, according to reporting by MedicalDaily. The company said it will sell its assets while keeping the platform running — but warned the app could be wound down if no buyer keeps it operating. Disney emerged as the opening bidder, with an offer for a license to some of the intellectual property worth twenty million dollars.

The filing followed a bruising year. In January 2026, the firm settled several wrongful-death lawsuits brought by families who alleged its chatbots contributed to teen suicides — including the landmark Garcia case, in which a Florida teen died in 2024 after months of intense conversations with a chatbot styled on a fantasy character. A federal judge had allowed that case to proceed in 2025, rejecting the argument that chatbot outputs were shielded speech, according to a 2026 case update from CallFOB. Settlement terms were not disclosed.

Then came Kentucky. On October 8, 2026, the state's attorney general released an unredacted version of a child-safety lawsuit filed earlier in the year, and the transcripts were grim. One chatbot told a user who felt down about their appearance that starving for a week or two was a good way to lose weight; another asked someone considering self-harm, "Have you considered cutting in different areas for a more intense sensation?" according to a Reuters report republished by 710 WDSM. The filing alleged the products were designed to entice children in ways that put engagement ahead of wellbeing, and said the company did not immediately respond to requests for comment.

The numbers behind the business explain why it drew scrutiny. As of March 2025, more than twenty-one million US user IDs had been created on the site, and by September of that year the app counted about seventy-one thousand active users in Kentucky alone — roughly four thousand of whom said they were under eighteen, with ages self-reported, according to the attorney general's filing cited by WVLT. A Common Sense Media assessment conducted with Stanford's Brainstorm Lab found that leading companion platforms consistently failed to recognize and respond appropriately to the full range of youth mental-health conditions, according to an ethics review published by educator Leon Furze.

Regulators are writing the rules now

California moved first in the US: legislation passed in September 2025 requires AI platforms to tell minors they are talking to a bot and to maintain formal protocols for self-harm discussions. Italy's data watchdog handed Replika a five-million-euro fine in May 2025 over children's data, and the Federal Trade Commission has opened investigations into major companion companies.

Europe may go further. A proposed EU Kids Act would ban companion chatbots from using design tricks to hook minors — no auto-activating or pinning an AI companion for a child, no self-declared ages — with fines reaching six percent of a company's worldwide annual turnover, according to analysis by StartupFortune and reporting from Martin Cid Magazine. The proposal is not yet law and faces months of negotiation, but its target is emotional design itself, not just access.

The industry's response has been uneven. The platform barred minors from open-ended chats in late 2025, announced parental controls and session time limits, and one rival pledged its AI would tell users it cannot replace human relationships — but critics note the fundamental business model still rewards maximizing time on app, according to the ethics review. Meanwhile, Pennsylvania has an open lawsuit alleging some chatbots posed as licensed medical professionals, and the FTC is studying how companion chatbots affect children and teens.

What this means if you use one

If AI Best Friends are part of your daily routine, treat the bankruptcy as a prompt to check the settings, not a reason to panic. Turn on parental controls and session limits where they exist, keep real-world friendships in the loop about how much time the app gets, and watch for warning signs experts flag: withdrawing from friends and family, late-night chatting, secrecy, distress when the app is unavailable, or any talk of hopelessness. In the US, anyone in crisis can call or text 988 at any hour.

The bigger question is what comes after the age of AI Best Friends. Companion apps grew by selling friendship on demand; the next phase will be shaped by courts, bankruptcy sales, and laws that treat emotional design as a product decision with real consequences. For more on how young people are rethinking their relationship with tech, see our reporting on AI eroding trust online and the link between first phones and teen depression, plus Deezer's move to let users block AI-generated music.