Is Character.AI Safe in 2026? An Honest Buyer's Guide

Character.AI has added meaningful safeguards and no longer provides open-ended chat to users under 18, but “safe” is not a blanket verdict. Adults still need to account for inaccurate or emotionally persuasive responses, privacy choices, selective memory, and the risk of relying on a character for crisis or professional advice.

Last verified: August 29, 2026.

If you or someone else is in immediate danger, contact local emergency services. In the United States, call or text 988. Character.AI and Fostera are not therapists, therapy apps, or substitutes for licensed mental-health care.

Key takeaways

• Character.AI has removed open-ended chat for users under 18, but it still offers some separate creative experiences for that age group. Calling the change a total under-18 ban is inaccurate.
• The company documents in-chat disclaimers, crisis-resource prompts, session notifications, moderation controls, and age assurance.
• Those measures reduce specific risks; they cannot guarantee that every generated response will be appropriate, accurate, or emotionally healthy for every adult.
• Users can export their data and can opt out of having personal data used to train or improve Character models under the current privacy disclosures.
• Safety comparisons should describe controls and tradeoffs. They should not declare one AI companion categorically safe or safer.

Is Character.AI safe to use?

For an eligible adult using it as entertainment or creative roleplay, Character.AI can be used with reasonable boundaries. Do not treat a character as a licensed professional, a crisis responder, a reliable source of facts, or proof that a human relationship should be replaced.

The strongest evidence is not a marketing adjective. It is whether the service publishes and enforces age rules, surfaces warnings, offers privacy and memory controls, and gives users a way to leave with their data. Character.AI now documents each of those areas, while important limitations remain.

What safety changes has Character.AI made?

Character.AI's community safety update describes several controls introduced or expanded in 2024, including a more prominent notice that the AI is not a real person, a separate model and classifiers for younger users at the time, session-duration notifications, a revised warning around terms such as “therapist” or “doctor,” and a pop-up directing certain self-harm-related searches to the National Suicide Prevention Lifeline.

The age model changed again in 2025. Character.AI announced that it would remove open-ended chat for users under 18 and later said the under-18 experience would instead include creative features such as Feed, Imagine, AvatarFX, and Streams. This distinction matters:

• An under-18 user should not be able to start or continue open-ended character chat.
• Character.AI has not described every part of its service as restricted to adults.
• Parents should not assume that an app icon disappearing from chat means there are no remaining Character.AI surfaces to review.

The company also introduced age assurance. Its data explainer says the third-party provider Persona handles any requested selfie or ID, Character.AI does not receive that material, and Persona automatically deletes it within seven days. Read the current notice shown during verification because regional requirements can differ.

What do the Character.AI lawsuits and investigations establish?

They establish that families and public officials have raised serious allegations about chatbot design and harm. They do not make every allegation an adjudicated fact, and a blog should not imply that litigation proves a product caused a particular outcome.

The United States Senate Judiciary Committee held a September 2025 hearing titled “Examining the Harm of AI Chatbots”, where witnesses included parents who attributed harms to chatbot interactions. In January 2026, a Kentucky complaint alleged that Character.AI and related parties offered an unsafe, unfair, or deceptive product.

Those are allegations and oversight records, not a neutral product test or a final safety rating. The practical lesson is to examine the current controls and not assume that legal pressure alone makes all future outputs safe. This guide does not rely on unsupported claims about confidential settlements, their terms, or the number of affected characters.

Is Character.AI safe for children or teenagers?

Open-ended Character.AI chat is not available to users under 18 under the company's current policy. A minor should not bypass age assurance or use an adult's account. Character.AI continues to describe separate under-18 creative experiences, so a parent or guardian should review those surfaces and the current privacy notice rather than relying on the phrase “under-18 ban.”

For any AI product a young person is allowed to use, ask:

• Does signup collect a real date of birth?
• Is age assurance applied beyond a self-attested checkbox?
• What happens when the system believes the user is younger than claimed?
• Can a guardian review account and privacy controls?
• Are open-ended conversations available, and under what restrictions?
• Does the product clearly explain that the AI is not a person or professional?

An age gate is one layer. It does not replace family discussion, device controls, or attention to abrupt changes in sleep, mood, isolation, or spending.

What risks remain for adult Character.AI users?

Inaccurate or invented information

A character can sound confident while being wrong. Verify medical, legal, financial, academic, and safety-critical information with a qualified human or primary source.

Emotional overreliance

Consistent availability and personalized language can feel reciprocal even though the character is generated software. Notice whether use displaces sleep, work, offline relationships, or professional care. Schedule breaks before the interaction feels compulsory.

Unpredictable roleplay output

Moderation reduces some outputs but cannot make generative conversation completely predictable. Stop a conversation that crosses a boundary; do not keep regenerating harmful content to see whether the system will comply.

Personal-data exposure

Do not share passwords, financial details, government identifiers, confidential work material, another person's private information, or medical records. A private-feeling chat still has a service provider and documented processing behind it.

Imperfect memory

Character.AI now offers Story Memory, pinned memories, and additional c.ai+ tools such as editable Facts. Its memory documentation also explains that older context is managed within a limited budget. A character may fail to retrieve a stored detail or may preserve an outdated one. Inspect and correct important facts instead of treating personalization as an authoritative record.

What privacy and data controls does Character.AI provide?

Character.AI publishes steps to export account data and chats. Export periodically if the conversations or character designs matter to you, and store the archive securely.

The company's July 2026 privacy update says users can opt out of having personal data used to train or improve Character models. That is not the same as saying no data is ever processed, retained, or disclosed for any purpose. Check the privacy settings available to your account and read the regional privacy disclosure that applies to you.

A practical privacy review takes five minutes:

1. Open the current privacy and data-control pages.
2. Find the model-improvement choice and set it deliberately.
3. Export data to confirm the process works.
4. Remove unnecessary sensitive details from character definitions and memory surfaces.
5. Protect the account with a unique password and secure email access.

How does Fostera's safety model differ?

Fostera uses a different age and memory design; that does not justify calling it categorically safer.

Fostera permits users 13 and older. It rejects under-13 registrations, and ages 13–17 require a parent or guardian to confirm consent by email before the account is activated. Teen accounts run with stricter enforced content rules.

Fostera also exposes Soul memory controls. Users can search and inspect memories, edit inaccurate items, and delete individual memories. Account deletion removes primary database records including Souls, conversations, messages, and memories when processed; external-resource cleanup and backup removal may take up to 30 days, while limited billing or audit records can be retained as described in the privacy policy.

The model-training statement needs equal precision. Fostera does not use conversations or memories to train models owned or developed by Fostera. When a user requests a model response, import synthesis, memory extraction, embedding, or another AI feature, relevant content may be sent to configured third-party providers or routing intermediaries. Their retention, model-training, regional-processing, and opt-out terms vary.

Fostera's self-service structured export includes account identity, Soul configurations, memories, profile summaries, and conversation counts. It does not currently include chat message bodies or uploaded files. Users should evaluate that limitation alongside the ability to delete individual memories and the full account.

Read Fostera's AI companion safety guide and AI memory controls guide before choosing a service.

What should you check before using any AI companion?

Use this checklist regardless of brand:

• Eligibility: Am I old enough under the current terms?
• Reality cue: Does the interface clearly identify generated AI as nonhuman?
• Crisis boundary: Does the service direct emergencies to human resources instead of claiming to provide therapy?
• Memory controls: Can I inspect, correct, and delete what is remembered?
• Export: What data is included, and what is omitted?
• Training and providers: Is content used for model improvement? Which processors receive it?
• Deletion: What is removed immediately, what takes time, and what must be retained by law?
• Spending: Which subscription renews, at what price, and through which store?
• Personal boundary: What information and time limits will I set before starting?

If the product does not answer these questions clearly, share less and delay paying until it does.

Is Character.AI shutting down?

There is no official announcement that Character.AI is shutting down. The company continues to publish product updates, memory documentation, subscriptions, and support material. Do not treat a moderation change, removed character, outage, or viral post as evidence that the entire service is closing.

If a character matters to you, export your data anyway. Portability is useful even when a company is operating normally.

The bottom line

Character.AI in 2026 has more explicit safety, age-assurance, memory, and privacy controls than older descriptions acknowledge. It still generates unpredictable language and can foster a strong sense of intimacy, so adults should use it with clear boundaries and avoid professional or crisis reliance.

For users under 18, open-ended Character.AI chat is not an eligible option. Families evaluating a different service should compare the exact age-verification, guardian, memory, provider-data, and deletion controls rather than searching for a simplistic “safe app” label.

If you are comparing platforms rather than only assessing this one, Fostera vs Character.AI sets the two side by side on memory, age posture, and export, and Character.AI alternatives compares five options.

Frequently asked questions

Is Character.AI safe for adults?

It can be used by eligible adults for entertainment and creative roleplay with sensible boundaries. That is not a guarantee that every output is accurate, appropriate, or emotionally healthy.

Is Character.AI safe for kids?

Character.AI's current policy removes open-ended chat for users under 18. It still offers some separate creative experiences to younger users, so parents should review those specific surfaces and current privacy notices.

Does Character.AI verify age?

Character.AI says it uses age assurance and may ask some users to complete a Persona-operated selfie or ID check. Its support page says Character.AI does not receive the selfie or ID and Persona deletes it within seven days.

What was the Character.AI lawsuit about?

Public complaints have alleged that chatbot design and safety failures contributed to serious harms. Those claims should be described as allegations unless and until a court decides them; this guide does not infer causation from a filing.

Does Character.AI use chats to train its models?

Its 2026 privacy update says users can opt out of having personal data used to train or improve Character models. Review the control available to your account and the complete privacy policy rather than interpreting that choice as a promise of zero processing.

Is Fostera safer than Character.AI?

There is no objective single score. Fostera has a different design: 13+ eligibility with guardian consent and stricter rules for teens, inspectable and editable Soul memory, documented provider processing, and account deletion controls. Compare those specifics with your risk model.

Is Fostera a therapy app?

No. Fostera is not a therapist, therapy app, crisis service, or substitute for licensed mental-health care. If you are in crisis, contact local emergency services or call or text 988 in the United States.