Character.AI Details Safety Updates for Support, Moderation and Teen Accounts

Character.AI’s latest safety update describes a shift from broad restrictions toward a wider set of product controls and support systems. The company says it continues to bar open-ended chat with Characters for users under 18, a decision it made last year, while expanding creative formats for teens and refining how it handles distress, moderation and age checks.

The update outlines product changes and work with outside organizations, but provides no performance figures for its detection or age estimation systems. That leaves a key question for users and developers: how well do the safeguards work in practice, and how often do they interrupt permitted activity?

Responding to distress across a conversation

Character.AI says its self-harm safeguards look at the context around a message, including signals that may accumulate over a long conversation. That design addresses a real limitation of single-turn detection: a message can be ambiguous on its own, while a pattern across multiple turns may indicate that a person needs support.

When the system identifies signs of distress, the company aims to respond with care and direct the user toward help outside the chat. It says clinicians and mental health experts have informed its approach, and it has worked with Koko, a nonprofit that provides free, self-guided emotional support tools.

The company also points users to resources from ThroughLine. Its directory covers crisis and mental health services in different countries, allowing referrals to be tailored by location instead of relying on one hotline number. The update gives no figures on how often these systems trigger, whether users follow the referrals or how the company evaluates whether a response is helpful. Those details would make it easier to understand the safeguards’ reach and limits.

Moderation decisions and user controls

Character.AI acknowledges that cautious moderation can affect content users consider acceptable. It says it is adjusting its processes to meet its safety goals while reducing mistaken actions. The practical change for creators is clearer notice: when content is moderated, they’ll be told what happened and why, and can appeal so the team can review the decision.

That gives creators a route to challenge a moderation outcome, though the update doesn’t specify review timelines, appeal volumes or how often an appeal changes the result. Those measures would help creators judge how the process works beyond the existence of an appeal option.

Users can also block another account and that account’s Characters, Posts, Voices and Scenes from search, discovery pages and profiles. Blocking works in both directions. Because the setting covers several content types and discovery surfaces, it may reduce unwanted encounters across the platform rather than only hiding a single profile. The company doesn’t describe whether blocked material can still appear through other routes.

Teen features and age assurance

Character.AI says many younger users came to the service for creative activities. It has continued developing the under-18 experience with formats including c.ai Series, Feed and Comics. The company presents these as ways to support creativity within the age-specific experience introduced after it removed open-ended Character chat for minors.

To direct users to the appropriate experience, Character.AI uses in-house age assurance technology and an age estimation model. It says it is working to improve accuracy and reduce disruptions for people who are 18 or older. The update doesn’t report accuracy rates, describe how users can resolve an incorrect estimate or explain what information the system uses. Those are important practical details because an age check can shape what features a person can access.

Parents and guardians whose email addresses are linked to a teen’s account can receive weekly activity updates through Parental Insights. Character.AI says it improved this tool with k-ID, which works on age-adaptive internet infrastructure. The update describes visibility into account activity, but doesn’t spell out which activity is included or how families can manage the email connection.

Outside expertise and open questions

The company says it will keep revising its policies and safeguards, and test changes before they reach users. It also names several organizations whose expertise informs its work. ConnectSafely, focused on online safety, privacy and digital wellbeing, has supplied teen council input and focus groups, as well as advice on educational materials.

Character.AI says its membership in the Internet Watch Foundation connects its detection work to a global effort to identify and prevent child sexual abuse material, and that it reports material it finds. It also belongs to StopNCII, which uses privacy-preserving digital fingerprints to find matching non-consensual intimate images without those images leaving victims’ devices.

The company’s update makes clear that safety depends on several layers: recognizing distress over time, offering routes to support, explaining moderation actions, giving users blocking controls and sorting accounts into age-specific experiences. Its next challenge is showing how those systems perform. Clear measures for detection quality, mistaken moderation, appeals and age estimation would help users assess whether the changes protect people while preserving the creative uses the platform is designed to support.