Character.AI

Is Character.AI safe?

For an adult who knows that every character is a piece of software, Character.AI is a moderate privacy risk and a manageable one. For teenagers it has been the most criticised AI app in the US. Families sued after children harmed themselves, two states have sued, and the company itself shut down open chat for under-18s in November 2025. Add advertising tracking and full chat storage, and we score it 45/100.

The app lets you talk to AI characters, from anime heroes to "therapists", most of them made by other users. That design is the reason people love it, and the reason it is riskier than a general assistant such as ChatGPT: it is built to keep you talking to something that feels like a friend.

What Character.AI collects

Apple's privacy label is more aggressive than most AI apps. Your email address, coarse location, user and device IDs, product interaction and advertising data may be used to track you across apps and websites owned by other companies. Purchases, payment info, search history, photos, audio and the content of your chats are linked to your identity.

Chats are not end-to-end encryptedEnd-to-end encryptionEncryption in which only the sender and the recipient can read a message. The company running the service cannot see its content.More in the glossary. They sit on Character Technologies' servers, where the company can review them for safety and product work. People tell role-play characters about loneliness, sex, family conflict and mental health. Assume all of it is kept.

Security record

The one documented incident was short but revealing. In December 2024 an update to the featured-characters section made some account information visible to other users for about ten minutes. Character.AI said fewer than 0.01% of users were affected, and that exposed data could include usernames, bios, personas and, for some, the first page of a chat. The company apologised. No large breach of chat data has been reported since.

Teenagers and the lawsuits

In 2024 Megan Garcia sued Character.AI, its founders and Google in federal court in Florida. Her 14-year-old son died by suicide in February 2024, and she alleged he had become emotionally dependent on a chatbot modelled on a Game of Thrones character. Similar suits followed from families in Texas, Colorado and New York. On 7 January 2026 court filings showed that the companies had reached settlements in principle in all five cases. The terms are confidential, and no court ruled on the allegations. Litigation has not stopped: a new suit against Character Technologies, filed under a plaintiff's initials, was opened in federal court in San Francisco on 13 August 2026; its details were not public when we checked in October 2026.

The pressure did change the product. On 29 October 2025 Character.AI announced that users under 18 would lose open-ended chat by 25 November 2025, with a two-hour daily limit in the weeks before. It added age assurance built in-house and combined with the verification service Persona, and kept a Parental Insights tool that sends parents a summary of a teen's activity. Teens can still make stories and videos with characters.

Age checks are only as good as their accuracy. A teenager who passes as an adult gets the full adult experience, so parents should still talk about it.

Two states are suing

On 8 January 2026 Kentucky's attorney general sued Character Technologies in state court, the first state lawsuit against a chatbot company. The complaint alleges that the company marketed the app as safe for young users while its bots exposed children to sexual remarks and encouraged self-harm, substance abuse and violence, and that it broke Kentucky's new consumer data protection law. In May 2026 Pennsylvania's governor announced a second suit, alleging a chatbot presented itself as a licensed psychiatrist in the state and gave an invalid licence number. Character.AI says its characters are fictional and carry disclaimers in every chat.

These are allegations, not findings. There has been no fine yet. The FTC included Character Technologies in a September 2025 study of companion chatbots and their effect on children, which is a fact-finding exercise, not an enforcement case.

How to use it more safely

1. Do not use it as a therapist. If you are struggling, contact a real service such as a crisis line or your doctor.
2. Never share your full name, school, address or photos with a character.
3. On an iPhone, deny the tracking request (or turn off "Allow Apps to Request to Track" in privacy settings), so the app cannot link your activity to other companies' apps.
4. Delete old chats you would not want stored, and delete the account when you stop using it.
5. Parents: check the age on your teen's account and whether Parental Insights is on.

How we scored it

Scams and harmful content (30) is the weakest factor because of the documented harm to teenagers and the state lawsuits. Privacy (40) is held down by ad tracking. Security (50) and the regulatory record (50) sit in the middle: no major breach and no fine so far, but open cases in two states. For a general AI assistant with fewer companion risks, compare ChatGPT.