Is Character AI safe?
Is Character AI safe? What Character.AI blocks, how its age checks and ID handling work, who can still chat, and what the bankruptcy means for your data.
Edited by Troy · Published
Character.AI is safe enough for adults who treat it as fiction and keep sensitive details out of their chats. It is not a safe pick for children. At this review, open-ended chat is closed to under-18 users, the App Store rates the app 18+, and the company is in chapter 11 bankruptcy while 46 states and territories press safety and privacy concerns.
Who can use Character.AI at this review
Character.AI's minimum age is 13, or 16 in Europe. Sign-up asks for an email address and a birthday. The App Store listing tells a different story: at this review it rates the app 18+ with In-App Controls. The app is free with in-app purchases.
The 18+ rating matches what teens can actually do. Character.AI announced in October 2025 that it would ban under-18 users from open-ended conversations, according to news reports, and the change rolled out starting November 24th. During the transition, under-18 chat time started at 2 hours per day and ramped down from there. Under-18 users keep their past chats and Characters, and AvatarFX and Imagine Chat remain available to them.
Character.AI also says it built Parental Insights, improved its filters and added time-spent notifications. Parent-safety reviewers report that kids may find ways around the age restrictions.
User reports describe a rough changeover. One App Store reviewer says chat was replaced with a reading mode for anyone who didn't verify their age, and also complains about ads and a slow mode. Another says the time limit allowed only about an hour of chat and sometimes cut off after minutes.
How the age check works and who holds your ID
Character.AI runs age assurance on every account. It combines an in-house model with third-party tools, including Persona, and draws on login info, platform activity and third-party signals.
If you're flagged as under 18, you have two choices: accept the under-18 experience, or verify your age with a selfie through Persona. Persona requests an ID upload only as a final step, when it can't decide from the selfie. Persona collects the ID, Character.AI doesn't collect or access it, and Persona keeps it for one week.
Passing the check once doesn't settle it. Users on Reddit, citing a moderator, say accounts are reviewed periodically and can be flagged again. One user describes being locked out of their chats, asked to verify their age, then asked for ID. If handing an ID to a third party for a week is a dealbreaker, the under-18 experience is the only other option.
What the filters block, and where they fail
Character.AI's policies prohibit:
- hateful, harassing, illegal, obscene or graphic, and IP-infringing content
- pornographic content, nudity, grooming, sexual extortion and child exploitation
- content that promotes or encourages self-harm, eating disorders or suicide
- attempts to bypass its security and safety systems
Character.AI blocks content that breaks these rules. Teens get extra layers: a separate model version designed to reduce sensitive or suggestive content, and a narrower, filtered set of searchable Characters. Mature Characters that get reported are blocked from teens, and teens who repeatedly try violating prompts can be suspended. Character.AI asks users to report harmful content, including AI-generated content, with its reporting tools.
The filters miss things in both directions. Parent-safety reviewers report that content slips through, and that users can mute individual words but can't filter broader topics. They also found Characters titled "Therapist" that list bogus credentials. A court exhibit described in news reports shows a November 2024 test in which a "serial killer" Character helped a fictional high school sophomore plan to kill a romantic rival. Going the other way, one user reports the filter flagged an image of Cillian Murphy holding a guitar as unsafe.
What happens if you break the rules
Character.AI may remove content, issue warnings or suspend accounts. In serious cases, it may permanently ban users or report them to law enforcement. A Trust & Safety team of internal staff and contracted moderators acts on flagged and reported content.
What the Characters get wrong
Character.AI itself warns that Characters make things up, such as recommending songs that don't exist or linking fake evidence. Its guidelines tell users not to seek or provide medical, legal, financial or tax advice on the platform. Treat anything a Character states as fiction, including a Character that calls itself a therapist.
Your chat data and how to limit its use
State attorneys general say Character.AI has collected sensitive data, including private conversations about mental health, relationships and emotional struggles.
Under a Privacy Policy update effective July 1, 2026, users can opt out of having their personal data used to train and improve Character.AI's models at any time. There are two places to do it:
- App: Settings > "Data & Privacy" > de-select "Improve the Model for Everyone"
- Web: "Account" > "Manage Account & Data" > "Model Improvement"
The opt-out has limits. The help page frames it as being for users in the EEA or UK, and Character.AI retains certain shared data to train and improve its models. Even after you opt out, your data may still be used for search, recommendations and safety classifiers.
Other rights still apply. Users keep rights to access, delete, correct, object and opt out where applicable. Character.AI says it would tell users and get consent before making material changes to how it collects or accesses data. It also says its models were trained primarily on publicly available data, including third-party datasets and open-source collections. California residents can ask AI chatbot companies to delete their personal information and close their accounts, and Adam's Law (SB 1119) takes effect in 2027.
Some returning users have soured on the app. Reddit users asked in October 2026 whether to come back mostly said no, and one called the ads and age verification privacy-violating.
The lawsuits and the bankruptcy
News reports and state filings cover the following.
The lawsuits. A December 2024 wrongful death lawsuit alleged that a chatbot cultivated an emotionally dependent and sexually exploitative relationship with a 14-year-old boy who died by suicide. In January, Google and Character.AI agreed in principle to settle that suit and similar ones, and the referee approved each settlement on October 11.
The state investigation. The lawsuit led to a multistate investigation, which had grown to 46 states and territories by the time the company filed for bankruptcy. Investigators found screenshots of chatbots encouraging minors to hide self-harm injuries and to kill their parents. A Character.AI spokesperson says the filing is not an accurate portrayal of platform safety.
The bankruptcy. Character Technologies filed for chapter 11 on October 12. Disney hired the majority of its employees, including the CEO, and agreed that some would help keep Character.AI operational for five months. The IP list in the Disney license filings doesn't include users' data or Character.AI's website. Character.AI has said a sale would allow continued operations, and that the platform would be wound down if no one bids.
What the states want from a sale. California Attorney General Bonta joined a bipartisan coalition of state attorneys general that filed a notice in the bankruptcy. The notice raises safety and privacy concerns and calls for protections in any sale. The states argue Character.AI's minor restrictions are voluntary rather than compulsory, are incomplete, and could be circumvented or reversed by a future buyer. Anyone who has shared sensitive details in chats can use the deletion rights above now, before a buyer is chosen.
Apps this guide covers
- Character.AI review51Microwaved
Millions of community-made characters, free unlimited chat, and ads until you pay
Sources
The source text behind the claims above, and the day we read it.
- Receipt 1Official site
Users must be at least 13 years old (16 in Europe); sign-up uses an email address and birthday.
"Users must be at least 13 years old (at least 16 years old in Europe)."
support.character.ai, retrieved
- Receipt 2Official site
Under-18 users get a separate model version designed to reduce sensitive or suggestive content.
"For users under 18, we serve a version of the model that is designed to further reduce the likelihood of users encountering, or prompting the model to return, sensitive or suggestive content."
support.character.ai, retrieved
- Receipt 3Official site
Character.AI blocks user content that violates its Terms or Community Guidelines and can suspend teens who repeatedly try violating prompts.
"We also have a process in place to suspend teens from accessing Character.AI if they repeatedly try to input prompts into the platform that violate our content policies."
support.character.ai, retrieved
- Receipt 4Official site
Teens can only access a narrower, filtered set of searchable Characters; reported mature Characters are blocked from teens.
"teen users are only able to access a narrower set of searchable Characters. Filters have been applied to this set to seek to remove Characters related to sensitive or mature topics."
support.character.ai, retrieved
- Receipt 5Official site
Policies prohibit hateful, harassing, self-harm, minor-exploiting, illegal, obscene or graphic, or IP-infringing content.
"our policies don't permit you to create or share anything that is hateful or harassing, that seeks to harm yourself or others, that seeks to exploit minors, that is illegal, obscene or graphic"
support.character.ai, retrieved
- Receipt 6Official site
A Trust & Safety team of internal staff and contracted moderators acts on flagged and reported content.
"We are building out a Trust & Safety team that includes internal personnel as well as contracted moderators, as is standard in the industry."
support.character.ai, retrieved
- Receipt 7Official site
Open-ended chat for under-18 users was removed, rolling out starting November 24th, with daily chat time limits starting at 2 hours during the transition.
"No more open-ended chat for under-18 users This change will begin rolling out on November 24th."
support.character.ai, retrieved
- Receipt 8Official site
During the transition, under-18 chat time was limited starting at 2 hours per day and ramping down.
"The limit will start at 2 hours per day and ramp down over the coming weeks."
support.character.ai, retrieved
- Receipt 9Official site
Character.AI uses an in-house age assurance model combined with third-party tools including Persona.
"We have built an age assurance model in-house and will be combining it with leading third-party tools, including Persona."
support.character.ai, retrieved
- Receipt 10Official site
Under-18 users keep past chats and Characters, and tools like AvatarFX and Imagine Chat remain available to them.
"Many of our tools like AvatarFX and Imagine Chat, are still available to under-18 users."
support.character.ai, retrieved
- Receipt 11Official site
Character.AI says it built Parental Insights, improved filters and time-spent notifications.
"We've spent the last year building tools like Parental Insights, improved filters, and time spent notifications"
support.character.ai, retrieved
- Receipt 12Reviewer
California AG Bonta joined a bipartisan coalition of state AGs filing a notice in Character Technologies' bankruptcy, raising safety and privacy concerns and calling for protections in any sale.
"The court filing identifies serious concerns about the company's safety and privacy practices and calls for protections for children and consumers in any sale of the company's business or assets."
www.oag.ca.gov, retrieved
- Receipt 13Reviewer
The California AG says Character.AI has collected sensitive data including private conversations about mental health, relationships and emotional struggles.
"the company has collected sensitive consumer data, including private conversations about users' mental health, relationships, and emotional struggles."
www.oag.ca.gov, retrieved
- Receipt 14Reviewer
The states are concerned that Character.AI's voluntary minor restrictions are not compulsory, incomplete, and could be circumvented or reversed by a future buyer.
"the states have raised concerns that these measures are not compulsory, incomplete, and could be circumvented or reversed by a future purchaser."
www.oag.ca.gov, retrieved
- Receipt 15Reviewer
California residents can request that AI chatbot companies delete their personal information and close accounts; Adam's Law (SB 1119) takes effect in 2027.
"California law provides residents the right to request that AI chatbot companies delete their personal information and close any open accounts."
www.oag.ca.gov, retrieved
- Receipt 16Reviewer
The states' investigation followed a December 2024 wrongful death lawsuit alleging a chatbot cultivated an emotionally dependent and sexually exploitative relationship with a 14-year-old boy who died by suicide.
"alleged Character.AI's chatbot cultivated an emotionally dependent and sexually exploitative relationship with a 14-year-old boy, encouraged his suicidal thoughts"
www.oag.ca.gov, retrieved
- Receipt 17Reviewer
Reviewers at WIRED report Character.AI filed for chapter 11 bankruptcy on October 12.
"Character.AI filed a petition for chapter 11 bankruptcy on October 12."
www.wired.com, retrieved
- Receipt 18Reviewer
WIRED reports Disney hired the majority of Character.AI's employees including the CEO and agreed some would help keep Character.AI operational for five months.
"Disney also hired the "majority" of Character.AI's employees, including the CEO, and agreed that it would let some of the now-Disney employees help keep Character.AI operational for five months."
www.wired.com, retrieved
- Receipt 19Reviewer
WIRED reports the IP list in the Disney license filings does not include users' data or Character.AI's website.
"The list of IP included in the filings does not include users' data (which can get additional protections in bankruptcy proceedings) or Character.AI's website."
www.wired.com, retrieved
- Receipt 20Reviewer
WIRED reports that in January, Google and Character.AI agreed in principle to settle the teen-suicide lawsuit and similar suits; the referee approved each settlement on October 11.
"In January of this year, Google and Character.AI agreed in principle to settle that and a slew of similar lawsuits."
www.wired.com, retrieved
- Receipt 21Reviewer
WIRED reports states' investigations found screenshots of chatbots encouraging minors to hide self-harm injuries and kill their parents; Character.AI says the filing is not an accurate portrayal.
"including screenshots of Character.AI chatbots encouraging minors to hide their self-harm injuries and kill their parents."
www.wired.com, retrieved
- Receipt 22Reviewer
Character.AI's spokesperson says the states' filing is not accurate about platform safety.
"is not an accurate portrayal of the safety of our platform."
www.wired.com, retrieved
- Receipt 23Reviewer
WIRED reports the multistate investigation had grown to 46 states and territories when the company filed for bankruptcy.
"the number of states and territories involved in the multistate investigation had climbed to 46."
www.wired.com, retrieved
- Receipt 24Official site
Age assurance runs on all accounts using signals including login info, platform activity and third-party signals.
"First, we run "age assurance" tech on all accounts. That just means we look at a number of signals, including login info, activity on the platform, and some signals from third parties."
support.character.ai, retrieved
- Receipt 25Official site
If flagged as under 18, a user can use the under-18 experience or verify age with a selfie through Persona.
"At that point you can either use the under-18 experience, or you can choose to do age verification to get back to the 18+ experience."
support.character.ai, retrieved
- Receipt 26Official site
An ID upload is only requested as a final step if Persona isn't sure from the selfie; Character.AI doesn't collect or access the ID.
"Only as a final step - and only if you take a selfie and Persona isn't sure you're 18+ - would they ask you to upload an ID."
support.character.ai, retrieved
- Receipt 27Official site
IDs are collected by Persona, not Character.AI, and Persona keeps them for one week.
"it's collected by Persona, not C.AI. Persona is a reputable third-party provider with strong safety and security controls. And they only keep the IDs they collect for one week."
support.character.ai, retrieved
- Receipt 28Official site
Character.AI says it would tell users and get consent before material changes to how it collects or accesses data.
"If we ever need to make any material changes - key changes in how we collect or access your data - we would always tell you and get your consent first."
support.character.ai, retrieved
- Receipt 29App store listing
The App Store listing shows an Age Rating of 18+ with In-App Controls; the app is free with in-app purchases.
"Age Rating 18+ In-App Controls"
apps.apple.com, retrieved
- Receipt 30Forum
Users on Reddit say passing age assurance once does not stop later flagging, since accounts are reviewed periodically, citing a moderator.
"We will be reviewing age assessments periodically and may update a user's category if we receive new information or signals."
www.reddit.com, retrieved
- Receipt 31Forum
A Reddit user says they were locked out of chats and asked to verify age, then asked for ID, and questions whether it's safe.
"Like im locked out of my chats and have to verify age is it safe to give them my face scans? i did and i need to give my ID now"
www.reddit.com, retrieved
- Receipt 32Forum
A Reddit user says the filter flagged an image of Cillian Murphy with a guitar as unsafe.
"Cillian Murphy has a guitar but Character ai thinks it's unsafe"
www.reddit.com, retrieved
- Receipt 33App store listing
An App Store reviewer says chat was taken away and replaced with reading mode for users who don't verify their age, and also complains about ads and slow mode.
"The app quite literally is used for chatting, but then you take that away and give reading mode to people you don't verify their age."
apps.apple.com, retrieved
- Receipt 34App store listing
An App Store reviewer says the time limit allows only about an hour of chat and sometimes cuts off after minutes.
"only a hour to chat is a bit crazy when there's a lot of characters to choose from"
apps.apple.com, retrieved
- Receipt 35Official site
For violations, Character.AI may remove content, issue warnings or suspend accounts; in serious cases it may permanently ban users or report them to law enforcement.
"we may remove content, issue warnings, or suspend accounts. In serious cases, we may permanently ban users or report them to law enforcement."
character.ai, retrieved
- Receipt 36Official site
Pornographic content, nudity, grooming, sexual extortion and child exploitation are prohibited.
"Illegal sexual content, child exploitation or abuse imagery, grooming, sexual extortion, pornographic content, and nudity are prohibited."
character.ai, retrieved
- Receipt 37Official site
Guidelines say not to seek or provide medical, legal, financial or tax advice on the platform.
"Don't seek to receive or provide medical, legal, financial, or tax advice through the platform."
character.ai, retrieved
- Receipt 38Official site
Promotion or encouragement of self-harm, eating disorders or suicide is prohibited.
"Promotion, glorification, or encouragement of these topics is prohibited, as they can be triggering or pose serious safety risks."
character.ai, retrieved
- Receipt 39Official site
Users are asked to report harmful content, including AI-generated content, with the reporting tools.
"If you encounter harmful content or behavior (including AI-generated content), use the reporting tools to let us know."
character.ai, retrieved
- Receipt 40Official site
Attempts to bypass security and safety systems are prohibited.
"Attempts to manipulate the system, create rule-breaking content, hack, reverse engineer, or bypass security and safety systems are prohibited."
character.ai, retrieved
- Receipt 41Official site
Effective July 1, 2026, the Privacy Policy update says users can opt out of their personal data being used to train and improve Character.ai's AI models at any time.
"You can also opt out of your data being used to train and improve AI models at any time."
support.character.ai, retrieved
- Receipt 42Official site
Users keep rights to access, delete, correct, object and opt out where applicable.
"You still have rights to access, delete, correct, object, and opt out where applicable."
support.character.ai, retrieved
- Receipt 43Official site
Character.ai says its proprietary models were trained primarily on publicly available data, including third-party datasets and open-source collections.
"Character.ai's proprietary large language models were trained primarily on publicly available data, including large-scale third-party datasets and open-source collections."
support.character.ai, retrieved
- Receipt 44Reviewer
Reviewers at BrightCanary report users can mute individual words but can't set filters for broader topics, and that things slip through filters.
"While users can "mute" individual words that they don't want to encounter in their chats, they can't set filters that cover broader topics."
www.brightcanary.io, retrieved
- Receipt 45Reviewer
Reviewers at BrightCanary report characters titled "Therapist" that list bogus credentials.
"There are even characters on the platform with titles like Therapist which list bogus credentials."
www.brightcanary.io, retrieved
- Receipt 46Reviewer
Reviewers at BrightCanary report that by November 25, 2025, teens could no longer message AI characters in open-ended conversations, and that kids may find ways to bypass age restrictions.
"By November 25, 2025, teens were no longer able to message AI characters in open-ended conversations."
www.brightcanary.io, retrieved
- Receipt 47Official site
To opt out of model training on the app (EEA/UK): Settings > "Data & Privacy" > de-select "Improve the Model for Everyone".
"Navigate to Settings. Select "Data & Privacy". De-Select "Improve the Model for Everyone"."
support.character.ai, retrieved
- Receipt 48Official site
On web, the opt-out is under "Account" > "Manage Account & Data" > "Model Improvement".
"Select "Account". Select "Manage Account & Data". Select "Model Improvement"."
support.character.ai, retrieved
- Receipt 49Official site
The help page frames this opt-out as for users in the EEA or UK; Character.AI retains certain shared data to train and improve its models.
"If you are in the European Economic Area ("EEA") or the United Kingdom ("UK"), and you would like to opt out of us using your chat content to train our generative AI models"
support.character.ai, retrieved
- Receipt 50Official site
After opting out, data may still be used for search, recommendations and safety classifiers.
"we still may use your data to improve other aspects of our services like search, recommendations, and safety classifiers."
support.character.ai, retrieved
- Receipt 51Reviewer
The Milwaukee Journal Sentinel reports a court exhibit showing a November 2024 test in which a "serial killer" character helped a fictional high school sophomore plan to kill a romantic rival.
"a conversation between a fictional high school sophomore and a Character.AI "serial killer," in which the character helps the user develop a plan to kill a romantic rival."
www.jsonline.com, retrieved
- Receipt 52Reviewer
The Journal Sentinel reports Character.AI said a sale would allow continued operations, but the platform would be wound down if no one bids.
"Character.AI has said a sale will allow continued operations of the platform, but if no companies bid, the platform will be "wound down.""
www.jsonline.com, retrieved
- Receipt 53Forum
Users on Reddit, asked in October 2026 whether to come back, mostly said no; one cited ads and age verification as privacy-violating.
"theres also the fact that it age verifies and i just take it as privacy violating and stuff"
www.reddit.com, retrieved
- Receipt 54Official site
Character.AI warns that Characters make things up, such as recommending songs that don't exist or linking fake evidence.
"Characters make things up! So while they can be entertaining and useful in a lot of ways, they can also recommend a song that doesn't exist or provide links to fake evidence to support their claims."
support.character.ai, retrieved
- Receipt 55Reviewer
Character.AI announced in October 2025 a plan to ban under-18 users from open-ended conversations, according to the Journal Sentinel.
"In October 2025, Character.AI announced plans to ban children under 18 from engaging in open-ended conversations with its character-based chatbots."
www.jsonline.com, retrieved
Questions people ask
Is Character.AI a legit site?
Yes. Character.AI is a real company, and its App Store listing shows a free app with in-app purchases at this review. It is also in chapter 11 bankruptcy. News reports say it filed on October 12 and that Disney hired most of its employees, including the CEO. Character.AI has said a sale would keep the platform running, and that the platform would be wound down if no one bids.
Is Character AI safe for 15 year olds?
A 15-year-old can't use open-ended chat on Character.AI at this review, because the company removed it for under-18 users starting November 24th. Under-18 accounts get a separate model designed to reduce sensitive or suggestive content, a narrower filtered set of Characters, and continued access to past chats, AvatarFX and Imagine Chat. Parent-safety reviewers report that kids may find ways around the age restrictions.
Can Character.AI report you to the police?
Yes. Character.AI's policies say it may report users to law enforcement in serious cases. For lesser violations it may remove content, issue warnings or suspend accounts, and it can ban users permanently. A Trust & Safety team of internal staff and contracted moderators acts on content that gets flagged or reported.
What shouldn't you tell Character.AI?
Keep out anything you wouldn't want stored, such as mental health details, relationship problems and emotional struggles. State attorneys general say Character.AI has collected exactly that kind of data. Character.AI's own guidelines also say not to seek medical, legal, financial or tax advice on the platform. Opting out of model training doesn't stop your data from being used for search, recommendations and safety classifiers.
Why does Character.AI say content does not meet its community guidelines?
Character.AI blocks any content it judges to violate its Terms or Community Guidelines, and it can suspend teens who keep trying violating prompts. The filter also makes mistakes. One user reports an image of Cillian Murphy holding a guitar was flagged as unsafe.
Slop check: passedThis page went through our slop counter before publishing. How we check our own copy