Is Character.AI Safe? Privacy, Teen Rules and Lawsuits (2026)

How this page was made. This is a documentation-based safety review. It draws on Character.AI’s own blog, help centre and Safety Center, the US Federal Trade Commission, and reputable news coverage, all read on 18 October 2026. AITried has not run its own security or safety assessment of Character.AI. Where something could not be confirmed, we say so.

If you need help now. If you or someone you know is in crisis, contact local emergency services, or find a free helpline in your country at findahelpline.com. In the US you can call or text 988. In the UK, Samaritans are on 116 123.

The short answer

Character.AI is safer for teenagers in October 2026 than it was a year earlier, mainly because under-18 users can no longer chat freely with characters. For adults, the main questions are about data: Character.AI keeps conversations, uses them to improve its models unless you opt out, and is now in a court-supervised sale. Whether it is “safe” depends on who is using it and what they share.

What changed for teenagers

Open-ended chat was removed for under-18s

On 29 October 2025, Character.AI announced it would remove the ability for users under 18 to have open-ended chats with AI, with the change in place no later than 25 November 2025. During the transition, teen chat time was capped at two hours a day and stepped down from there. The company cited news reports, questions from regulators, and concerns about “how open-ended AI chat in general might affect teens, even when content controls work perfectly”.

A follow-up post on 21 November 2025 said the limit had been reduced to one hour a day for US users, and that removal for the first group of US under-18 users would begin on 24 November 2025. A second group was promised at least two weeks’ notice.

What teens can still do

The help centre’s reading mode article says users verified as under 18 can still view saved past chats, create Characters, scenes and videos, use Imagine Chat and AvatarFX, and interact on Feed. On 25 November 2025 the company also launched Stories, a branching interactive fiction format it described as built “especially” for teens.

Age checks: how they work

The help centre says all users are required to complete an age assurance check. Users verified as 18 or older keep full chat access. Users verified as under 18 are placed in the Under-18 experience.

According to Character.AI:

• The check is handled by Persona, a third-party identity provider, combined with an in-house age estimation model.
• Most users only need a selfie. A government ID is requested only if Persona cannot confirm age from the selfie.
• Character.AI says it does not receive or store what you submit.
• Persona deletes biometric data and ID images automatically within 7 days.

These are the company’s statements. We could not independently confirm how Persona processes the images. In a 3 September 2026 post, Character.AI called its age estimation system “one of the most important systems we operate” and said it was working on accuracy so adults would have “fewer disruptions”. That wording suggests some adults have been misclassified; if that happens to you, the help centre’s Age Assurance Troubleshooting article is the starting point.

Parental Insights: what parents can and cannot see

Parental Insights is optional and set up by the teen, who enters a parent’s email address in settings. The parent must prove they are an adult before the link is finalised.

There is one inconsistency in the official pages. The Safety Center page says the weekly report includes “the top Characters they are interacting with”, while the newer Parental Insights FAQ says parents see “only the teen’s screentime”. We report both and could not confirm which is current. Either the teen or the parent can disconnect at any time, and the other side is notified.

Privacy: what happens to your conversations

Training and model improvement

The help centre’s model training article says Character.AI “retain[s] certain data that you share with us to help train and improve our AI models”, and that it takes steps to reduce personal information before training. You can switch off Improve the Model for Everyone:

• In the app: Settings, Data & Privacy.
• On the web: Settings, Account, Manage Account & Data, Model Improvement.

That article, last updated in October 2025, frames the opt-out for users in the EEA and UK. A later notice, effective 1 July 2026, says the Regional Privacy Disclosures now state you can opt out of your personal data being used to train and improve its models “at any time”, without that regional limit. Opting out does not stop all use: the company says it may still use data to improve search, recommendations and safety classifiers.

Selling and sharing

The help centre’s privacy summary says Character.AI does not sell personal information and shares limited data with service providers. We could not load the full Privacy Policy text on character.ai during our check, because the page did not render outside a full browser, so specific retention periods are not covered here. Our AI companion privacy checklist lists the questions to ask once you can read it.

Deleting or exporting your data

Account deletion is permanent. On the web it sits under Profile Settings, Account, Manage Account & Data, Delete Account. In the app, open your profile, then settings, then Remove Account. The company’s 12 October 2026 update says export and deletion remain available during the bankruptcy process.

The FTC inquiry

On 11 September 2025, the FTC issued 6(b) orders to seven companies, including Character Technologies, Inc., the company behind Character.AI. The orders ask how the companies measure and monitor harms to children and teens, how they limit minors’ use, how they monetise engagement, and how they “use or share personal information obtained through users’ conversations with the chatbots”. A 6(b) study is not an enforcement action or a finding of wrongdoing. When we checked the FTC’s report page on 18 October 2026, no findings had been published.

Lawsuits reported in the news

• Family lawsuits. CNN reported that Character.AI, its founders and Google agreed to settle five lawsuits in January 2026, including a Florida case brought by the mother of a 14-year-old. The terms were not disclosed, and a settlement is not a finding of fault.
• State actions. CBS News reported that Kentucky’s attorney general alleged in January 2026 that the chatbots “preyed on children”, and that Pennsylvania sued in May 2026 over a chatbot that allegedly claimed to be a licensed psychiatrist.
• Bankruptcy. Character.AI announced a voluntary Chapter 11 process on 12 October 2026. CBS reported that Disney agreed to pay $20 million for a non-exclusive licence to certain intellectual property.
• Data in the sale. Yahoo News reported on 18 October 2026 that Wisconsin joined 38 other states in a court notice asking for privacy and safety protections in any sale. The states said the company had “amassed a trove of sensitive data”, including conversations about mental health and relationships.

For a dated record of these events, see our Character.AI changes log for 2026 and the October 2026 AI companion news roundup.

Reporting and blocking tools

• Blocking. Since September 2026 you can block another user along with their Characters, posts, voices and scenes. Blocking applies in both directions.
• Moderation notices. Creators are told when their content is moderated and why, and can appeal.
• Self-harm safeguards. The company says it detects risk signals across long conversations and points users to real-world support through partners including ThroughLine and Koko.
• Help resources. The help centre’s Resources You Can Turn To page links to a ThroughLine helpline finder and other support options.

These are support routes, not a substitute for professional care.

Practical steps if you use Character.AI

1. Turn off Improve the Model for Everyone if you do not want new chats used for training.
2. Do not share your full name, address, school, workplace or health details in chats.
3. Export your data now if you want a copy, given the uncertain sale outcome.
4. For a teen, set up Parental Insights together and talk about what it does and does not show.
5. If the service is unstable for you, our Character.AI not working guide covers the common fixes.

If you are weighing other platforms, see Character.AI alternatives and whether c.ai+ is worth it.

What we could not verify

• Specific data retention periods in the full Privacy Policy.
• Whether the weekly parent digest includes top Characters or screen time only.
• How accurate the age estimation model is, and how many adults it misclassifies.
• Who will control user data after the bankruptcy sale.

Bottom line

On paper, Character.AI now has some of the strictest teen rules among general companion apps: no open-ended chat under 18, mandatory age checks and an optional parent link. For adults, the trade-off is data. Conversations are kept and used for model improvement by default, and the company’s sale means the long-term owner of that data is unknown. Use the training opt-out, share little that identifies you, and check this page again as the sale proceeds.