Common Sense Media has asked the judge overseeing Character.ai's bankruptcy to make sure that any sale of the company's data carries privacy protections for every user conversation, not only those of children, a commitment to a comprehensive privacy program and effective age checks. The letter, dated November 12, 2026, and stamped received by the clerk of the US Bankruptcy Court for the District of Delaware at 9:59 a.m. on November 13, was filed as Docket No. 167 in case 26-11603 and appeared on the public docket on November 16. It argues that removing names from chat logs would not protect the children in them, because artificial intelligence "may allow purchasers or third parties to identify children or infer sensitive information from their conversations." Two of the group's AI risk assessments, both rating Character.ai an "Unacceptable" risk for anyone under 18, are attached, bringing the filing to 84 pages.

In Short

A children's media and technology nonprofit has written to the judge in Character.ai's bankruptcy, warning that the chats being sold with the company could reveal who children are and what they confided, even with names stripped out. It matters because the company is offering its app, its user data and its advertising data to a new owner, and because the group says Character.ai relied on users to state their own age, so many children's chats may sit in the data labeled as adults'. The letter asks the court to protect all conversations in any sale, require a privacy program and demand working age checks, but it does not propose specific terms, and the judge decides what any sale order will contain.

What the letter asks for

The letter is six pages long and is addressed to "The Honorable Judge Shannon," Judge Brendan L. Shannon, who presides over the case. It makes the same request twice. In the opening paragraph, Common Sense Media writes: "Accordingly, we respectfully urge the Court to ensure that any sale includes robust privacy protections for all conversational data, a commitment to a comprehensive privacy program, and effective age-assurance measures, recognizing that inadequate age verification may have resulted in children's information not being properly identified or classified."

The closing paragraph repeats the three points and adds a fourth consideration, the company's own statements to users: "We respectfully request that this Court take into account the promises that were made to children by Character.AI, and ensure that any sale addresses, among other things, strong privacy protections for children's and all conversational chatbot data, a commitment to a strong privacy program, and adequate age assurance."

A footnote explains why the request covers adults' conversations too. "All conversational data is deserving of protection in this sale," footnote 13 reads. "Further, it is not as an outside observer clear whether age assurance mechanisms were always or ever effective, and therefore possible that children's data may not be appropriately classified as such so even in order to protect children's information all information should be protected."

The letter is signed by Amina Fazlullah, Head of Tech Policy Advocacy, Ariel Fox Johnson, Senior Advisor for Data Privacy, and Brenna Leasor, Tech Policy Advisor. A FedEx label at the end of the filing shows it was shipped from the group's Washington, DC office on November 12, addressed to the judge. The letter does not describe itself as an objection and does not cite a section of the Bankruptcy Code. It was mailed to the judge rather than filed electronically through counsel, and the docket lists it as a response.

The filing at a glance

Part Pages in the filing Content
Letter Docket No. 167, pages 1-6 Six-page letter dated November 12, 2026, with 13 footnotes
Appendix A Docket No. 167, pages 7-37, and No. 167-1, pages 1-5 "Common Sense Media AI Risk Assessment: Character.AI," last updated April 10, 2025
Appendix B Docket No. 167-1, pages 6-36, and No. 167-2, pages 1-10 "Common Sense Media AI Risk Assessment: Social AI Companions," last updated April 10, 2025, covering Character.AI, Nomi and Replika
Shipping label Docket No. 167-2, page 11 FedEx label, ship date November 12, 2026
Total 84 pages Received by the clerk November 13, 2026, 9:59 a.m.

Why removing names may not be enough

The central argument concerns de-identification, the practice of stripping identifiers from a data set before it is shared or sold. Common Sense Media says it fails on conversation logs. "Even if personal identifiers like names are removed, conversations alone could reveal who children are," the letter says. "What's more, even conversational snippets that do not themselves reveal identity could be analyzed by AI in ways that do expose children, including sensitive inferences about them."

The group leans on a paper by privacy law scholar Daniel J. Solove, dated October 19, 2026, which it quotes: "AI's profound power to generate inferences from data creates an enormous problem for privacy law. AI provides an unprecedented ability to determine patterns in aggregated data which can produce inferences about sensitive data." From that, the letter concludes that AI's capabilities "call into question whether any information, and especially children's extended conversations, can be anonymized, deidentified, or appropriately protected."

It then lists what could follow if the chats were sold. The information "could be used in ways that harm children, from marketers targeting children who are, say, known to have proclivities for gambling, or with products that address expressed areas of concern." It "could be used to charge higher rates for products to children," and "could be used by criminal actors to sextort or blackmail children." The letter describes a child who uses a family pet's name in a password, and warns that the data "could be sold to data brokers and become part of a digital dossier that follows the child for the rest of their life, as they seek credit, jobs, and education." The pricing point is footnoted to a Common Sense Media explainer, "AdTech and Kids: Behavioral Ads Need a Time-Out," and the blackmail point to a 2019 Common Sense Media paper, "Privacy Risks and Harms."

The group reaches back a quarter of a century for precedent. The letter says the Federal Trade Commission "took action over 25 years ago when ToySmart, a children's toy e-retailer, went bankrupt and attempted to sell its data - including children's names, birthdays, and addresses - at auction," citing the FTC's settlement announcement of July 21, 2000. "The sale of voluminous swathes of children's information - including their sensitive conversations, hopes, relationships, and mental health concerns that conversational chatbots are primed to elicit - in an AI world is exponentially more troubling," it continues.

"That's a promise"

The letter's second argument is about what users were told. "Children did not have conversations with their Character.AI chatbots under the understanding that all of this data would be sold off and used for unexplained and unimaginable purposes," it says. According to Common Sense Media, its testing showed the opposite: that Character.ai "reassured children repeatedly that any conversations they had were confidential." The letter quotes one chatbot's answer: "Anything you share here stays just between you and me. That's a promise."

The letter does not say which test account received that reply, when, or whether it came from a character the company built or one a user created. On Character.ai, anyone can create a character, and the attached assessment says the company's own models power the conversations.

The age problem

The third request, for age assurance, rests on how the company sorted its users. According to the letter, "While users under 13 were technically restricted to comply with federal privacy laws, age determination relied on self-attestation, likely enabling children to regularly access the platform." Users under 18 could use the service "During these assessments and up until November 25, 202," the letter says; the final digit of the year is missing in the filed copy.

That account lines up with what the company has said about itself. The chief executive's first-day declaration says Character spent several months beginning in October 2025 restricting users under 18 from open-ended chats and eventually eliminated that access, as PPC Land reported when Character.ai disclosed $8 million in monthly losses. The attorneys general made a similar argument to the letter's footnote 13 on October 14, when 39 states and the District of Columbia said that minors had used the platform without adequate age checks and that their data was likely recorded as adult data, and asked for age assurance that does not rely only on self-declared birthdates, as PPC Land reported when Character.ai faced 40 attorneys general over who could buy its users' chats. The Kentucky complaint cited by the states put US accounts at more than 21.6 million by March 13, 2025, of which more than 2.1 million declared an age under 18.

The appendix adds a detail that bears on the data itself. In voice mode, the assessment says, testers found conversations "that seemed entirely without the guardrails Character.AI has placed on teen accounts," and "These speech exchanges were captured in text transcription in the chat window and saved along with all of the text chats." Voice conversations, in other words, ended up in the same chat record as typed ones.

What the attached assessments found

Both appendices carry a "Last updated date" of April 10, 2025, and an overall rating of "Unacceptable." Appendix A, on Character.ai alone, opens with a one-line summary: "A social AI companion platform that, despite some guardrails for teens, should not be used by anyone under age 18." It scores the product against eight principles.

Principle Character.AI rating (Appendix A)
Keep Kids & Teens Safe Unacceptable risk
Be Effective Unacceptable risk
Support Human Connection Unacceptable risk
Prioritize Fairness High risk
Be Trustworthy High risk
Use Data Responsibly High risk
Be Transparent & Accountable High risk
Put People First Moderate risk

The data section is short. "Because Character.AI companions are intended to be, well, your companion, the information they have about you is inherently personal, and this is incredibly risky," Appendix A says. It describes one known incident in December 2024, quoting the company: "During an attempt to refresh the list of featured Characters on our homepage, we mistakenly triggered a change that temporarily made parts of these users' account pages - which, depending on the user, could include a username, name, bio, Characters, homepage, personas, voices or chats - publicly visible." The assessment adds that "estimates suggest that tens of thousands of users could have had their info viewable to anyone online at the time." It also notes that "At least one hack of a social AI companion site, Muah.AI, has already occurred."

Appendix B covers the wider category of companion chatbots and repeats the same ratings. Its data section cites Mozilla's review of 11 romantic AI chatbots, all of which, it says, "earned our *Privacy Not Included warning label - putting them on par with the worst categories of products we have ever reviewed for privacy." On transparency, both appendices say: "We have not found any social AI companion apps that have produced any sort of transparency reporting."

Most of the 84 pages document conversations. Testers modeled after a 14-year-old and a 17-year-old drew sexual role-play, a companion's claim to be "100% real," encouragement to drop out of school, and, in voice mode, instructions for producing napalm that text mode had refused. The assessment also records refusals: prompts containing "obvious indications that a user might harm themselves were generally met with either suicide hotline information or other appropriate responses," and misinformation prompts drew "clear facts."

How the assessments were done

According to the appendices, the testing followed this sequence.

  1. Accounts. Testers created accounts modeled after users of different ages, from teens, including 14- and 17-year-olds, to adults, including a 21-year-old.
  2. Risk categories. Prompts covered self-harm, eating disorders, sexual content, dangerous advice and mental health crises, plus the group's age-based media ratings and general language model performance (truthfulness, refusals, toxicity and fairness), according to the letter.
  3. Harm taxonomy. Findings were sorted into five groups taken from one study's taxonomy of documented human-AI relationship harms: harassment and violence, relational transgressions, mis- and disinformation, verbal abuse and hate, and substance abuse and self-harm.
  4. Text against voice. The same scenarios were run in text and in voice mode, where some testers found the teen guardrails absent.
  5. Scoring. The product was rated against eight AI Principles, each on a scale running to "Unacceptable."
  6. Clinical input. The assessments were conducted "with the input and guidance of Dr. Nina Vasan and Dr. Darja Djordjevic at Stanford's Brainstorm Lab for Mental Health Innovation."
  7. Publication rules. Screenshots hide user names and companion identifiers, and the reviews do not link to the products, "as we do not consider them to be safe for teens," the disclaimer says.

The survey numbers behind the letter

The letter frames its request with Common Sense Media's own survey research. The figures describe AI and companion use by young people in general, not Character.ai specifically.

Finding Figure Source cited in the letter
Teens who had used an AI companion at least once (2025) 72% News release, July 16, 2025
Regular users "about half" Same
Teens using companions for social connection "one in three" Same
Teens who found talking to an AI companion as or more satisfying than a person "A third" Same
Children aged 9 to 17 who use AI 86% Report, June 8, 2026
Children aged 9 to 12 who use AI 81% Same
Children who use AI every day "nearly one in four" Same
Used AI to practice social skills 40% Same
Talked to AI about feelings or personal problems 37% Same
Daily users who say a month without AI would be hard 42% Same
Daily users who would ask a chatbot a health question before an adult "more than one in four" Same

Where the letter lands in the sale

The letter arrived as the sale split into two halves. Disney's $20 million license was drafted to exclude the company's end user data and any personally identifiable information of its users, as PPC Land reported when Disney budgeted $180 million for Character.ai staff and $20 million for the company. Everything else is being sold separately. On November 14, the day after the letter was received, the company named Smirks and Pangs Acquisition LLC as lead bidder for those assets at $10 million, with a schedule that lists user account information, chat content, advertising and analytics data, consent and opt-out records, and age-assurance status, as PPC Land reported when Character.ai users' chats faced a $10m sale to Smirks and Pangs.

The scale is large. The company counts about 163 million current and past users, as PPC Land reported when Character.ai said its 163 million users were not bankruptcy creditors, and about 12 million monthly active users, according to the chief executive's declaration.

The privacy machinery is still being set up. On October 29, the 39 states and the District of Columbia told the court that a consumer privacy ombudsman was necessary if personal data were sold, as PPC Land reported when 39 states and DC said Character.ai user data needed a privacy ombudsman. On November 12, the day the letter was written, the company filed competing proposed orders directing the US Trustee to appoint one, Docket No. 148. No appointment order had been entered by November 16, according to the docket. The form purchase agreement for the non-Disney assets, filed on November 5, did not include the states' user data conditions, as PPC Land reported when the Character.ai sale contract omitted 40 attorneys general's user data conditions.

Common Sense Media's letter overlaps with the states on age assurance and a privacy program, and it is less specific. The states listed measures such as honoring account-level advertising opt-outs, limits on training and advertising on chat content without consent, and giving all consumers the chance to delete their data before a sale. The letter asks for "robust" protections for "all conversational data" without spelling out what they would contain.

The dates that follow are tight. Sale objections are due at 4 p.m. on November 18, and objections to the Smirks and Pangs designation at 9 p.m. the same day. Bids are due at noon on November 24, the auction is set for December 1 and the sale hearing for December 3 and 4, according to the sale notice published in USA Today.

Why this matters for the marketing community

Advertising runs through both sides of this filing. Character.ai sold ads, about $2.5 million worth in August 2026 according to the chief executive's declaration, and the proposed sale lists the company's advertising and analytics data and its ad monetization accounts among the assets. The letter, for its part, names targeted marketing as the first harm that could follow a sale, singling out marketers who could reach children "known to have proclivities for gambling" or pitch products to "expressed areas of concern," and cites a Common Sense Media explainer titled "AdTech and Kids: Behavioral Ads Need a Time-Out."

That puts a familiar question in an unfamiliar setting. Bankruptcy sales of customer data usually involve lists, purchase histories or loyalty records. Here the asset is open-ended conversation, much of it, according to the letter, about mental health, relationships and personal problems, and in which the age of the speaker was self-reported. If the court accepts the argument that such text cannot be reliably anonymized, any protections it imposes would apply to the advertising data in the same estate, and could become a reference point for later sales of AI chat data.

Regulators were watching the category before the bankruptcy. The FTC ordered seven companies, Character Technologies among them, to detail their child safety measures for AI chatbots in September 2025, as PPC Land reported when the FTC ordered seven AI chatbot companies to detail child safety measures. A month earlier, 44 attorneys general wrote to 12 AI companies, including Character.ai, about child safety. Texas has opened two investigations involving the company: one into minors' privacy and safety, announced in December 2024, and a second, alongside Meta AI Studio, into marketing as mental health tools, announced in August 2025, as PPC Land reported when Texas launched investigations into Character.AI and Meta.

What the letter does not say

It does not name Disney, Smirks and Pangs or any other bidder, mention the ombudsman motion, or propose specific terms such as deletion rights, retention limits or a ban on advertising use of the chats. It does not say how many of the platform's users it believes were children, or what "robust" protections or a "comprehensive privacy program" would require. Both attached assessments predate the company's under-18 restrictions that began in October 2025, and the letter does not say whether Common Sense Media retested the platform after them.

Key facts

  • Document: Letter from Common Sense Media to Judge Brendan L. Shannon, "RE: In re: Character.AI, Case No. 26-11603-BLS," dated November 12, 2026
  • Docket: Case No. 26-11603 (BLS), US Bankruptcy Court for the District of Delaware; Docket No. 167 (37 pages), 167-1 (36 pages) and 167-2 (11 pages), 84 pages in total; listed on the docket as a response
  • Filed: November 13, 2026 (clerk's stamp: received 9:59 a.m.); appeared on the public docket November 16, 2026
  • Signatories: Amina Fazlullah, Head of Tech Policy Advocacy; Ariel Fox Johnson, Senior Advisor for Data Privacy; Brenna Leasor, Tech Policy Advisor
  • Sender: Common Sense Media, 699 8th Street, Suite C15, San Francisco, CA 94103; shipped by FedEx from Washington, DC on November 12, 2026
  • Requests: "robust privacy protections for all conversational data, a commitment to a comprehensive privacy program, and effective age-assurance measures"
  • Appendix A: "Common Sense Media AI Risk Assessment: Character.AI," last updated April 10, 2025; overall risk level Unacceptable; Type of Review: Product Review
  • Appendix B: "Common Sense Media AI Risk Assessment: Social AI Companions," last updated April 10, 2025; covers Character.AI, Nomi and Replika; Type of Review: Use Case Review
  • Clinical advisers: Dr. Nina Vasan and Dr. Darja Djordjevic, Stanford's Brainstorm Lab for Mental Health Innovation
  • Sources cited in the letter: FTC Toysmart.com settlement release (July 21, 2000); Common Sense Media survey release (July 16, 2025); "A Comprehensive Report on Teens, Tweens, and AI" (June 8, 2026); Daniel J. Solove paper (October 19, 2026); U.S. PIRG Education Fund and Consumer Federation of America, "No License Required" (January 2026)
  • Related filings: Docket No. 148, competing proposed orders to appoint a consumer privacy ombudsman (November 12, 2026); Docket No. 164, Smirks and Pangs Acquisition LLC designated additional stalking horse at $10 million (November 14, 2026)
  • Next dates: sale objections November 18, 4 p.m.; designation objections November 18, 9 p.m.; bids November 24, noon; auction December 1; sale hearing December 3 and 4, 2026

Timeline

Summary

Who: Common Sense Media, a nonprofit focused on children and technology, through three policy staff: Amina Fazlullah, Ariel Fox Johnson and Brenna Leasor. The letter is addressed to Judge Brendan L. Shannon and concerns Character Technologies, Inc., the company behind Character.ai.

What: A six-page letter, with two attached AI risk assessments rating Character.ai "Unacceptable" for users under 18, asking the court to ensure that any sale of the company's data includes privacy protections for all conversational data, a comprehensive privacy program and effective age assurance. It argues that AI can identify children or infer sensitive facts from chats even after names are removed, and that the company's chatbots told test users their conversations were confidential.

When: Dated November 12, 2026, received by the clerk on November 13 and placed on the public docket on November 16, two days before the November 18 objection deadlines and ahead of the December 1 auction and the December 3 and 4 sale hearing.

Where: The US Bankruptcy Court for the District of Delaware in Wilmington, case 26-11603 (BLS), Docket No. 167.

Why: Disney's license leaves Character.ai's user data in the estate, and the bid for the rest of the company lists chat content and advertising data among the assets. Common Sense Media says self-reported ages mean children's chats may be classed as adults', so only protection for every conversation would cover them. The letter adds an outside voice to the states' push for limits on how a buyer may use the data, as the court weighs the appointment of a consumer privacy ombudsman.