Can Character AI actually see my chats?

Yes. Character AI's own privacy policy lists "chat communications" among the content it collects and states that this data is used to "train our artificial intelligence/machine learning models." Every message you type is stored and used for training.

Does Character AI share my conversations with anyone else?

It can. According to the privacy policy , chats may be shared with affiliates, vendors performing content moderation, analytics providers, and law enforcement responding to a warrant, subpoena, or court order.

Can I opt out of Character AI training on my chats?

No — unlike OpenAI, which lets users disable training on their content, Character AI's policy offers no such control. Independent reviewers note it stores every message and uses all of it , with no opt-out and no encryption commitments stated in the policy.

How long does Character AI keep my chat history?

The policy doesn't say. It only states data is kept "as long as necessary," with no defined retention limit — a phrase with no clock on it . By contrast, OpenAI processes deletion requests within 30 days.

Do human employees read my private Character AI chats?

Neither the official policy nor the support FAQ states whether staff review individual conversations — that gap is worth noticing. What's confirmed is that chats are collected, used for training, and may be disclosed to moderation vendors and law enforcement.

How can I protect my privacy when chatting with AI companions?

Never share sensitive personal information, use privacy controls like temporary chats and training opt-outs where available, and re-read policies regularly since terms change. Stanford's Dr. Jennifer King advises avoiding uploading sensitive files without redacting personal details first.

Can Character AI See Your Chats? Privacy Risks Revealed

The Short Answer: Yes, Character AI Can See Your Chats

Yes — Character AI can see your chats, and the company says so itself. Its privacy policy explicitly lists "chat communications" among the content it collects, and states that collected information is used to "train our artificial intelligence/machine learning models." If you've been treating your conversations with a Character as private, the policy says otherwise.

The details matter. According to Character AI's own privacy policy, chats may be shared with affiliates, vendors and service providers — including those performing content moderation — analytics providers, and law enforcement responding to a warrant, subpoena, or court order. Voice recordings are also collected if you use voice features, alongside identifiers like your User ID, Chat ID, and Session ID.

The policy even warns users directly: "Sensitive personal information is not required or necessary to use the Services – please do not include any sensitive personal information in your interactions with us or the Services." In other words, the company acknowledges the risk and asks you to self-censor — while every message you type is stored and used for model training. Independent reviews put it bluntly: Character AI "stores every message and uses all of it," with no opt-out and no encryption commitments stated in the policy.

What that means in practice:

No opt-out from training. Unlike OpenAI, which lets users disable training on their content, Character AI's policy offers no such control.
No defined retention limit. The policy does not specify how long chats are kept, only that data is retained "as long as necessary."
A broad content license. Character AI's Terms of Service grant the company a "broad, perpetual license to exploit and commercialize user content."
Third-party disclosure. Your chats may reach content moderation vendors, analytics providers, and law enforcement.

One honest caveat: neither the official policy nor Character AI's support FAQ states whether human staff read individual private chats. What's confirmed is that chats are collected, used for training, and may be disclosed — the exact mechanics of who sees what remain unspecified.

The context makes this more than a technicality. Character AI faces an FTC probe, lawsuits and investigations from multiple state attorneys general, and a 42-state warning letter, and it has been barred from allowing under-18 users in open-ended chats. It's also been rated F for safety (22/100) in one independent review and C+ for privacy in another 2026 privacy guide — of 15 AI companion apps reviewed there, only 2-3 handled user data responsibly.

This is the uncomfortable reality of AI companionship today: the most intimate conversations — the ones you'd never post publicly — are often treated as training data. One reviewer wrote that "the idea that some unknown company could be reading an 'anonymized' version of those conversations made my stomach turn." That's why we built FlirtAlly around a different standard: honest safety, with a PG/Flirt default experience, ID-verified opt-in access to the 18+ Spicy side, hard-coded prohibited categories, and in-app reporting — stated plainly, not hidden in fine print.

If you're choosing a companion platform, this is the question to ask first: who can see your chats, and what happens to them after you hit send?

What the Policy Really Says: No Opt-Out, No Retention Limit, Broad License

Most people never read past Character AI's friendly FAQ — and that's exactly where the story changes. The official support page frames data use as "service improvement" and highlights your access, deletion, and portability rights, but the underlying policy tells a sharper one.

Character AI collects "chat communications" as user-submitted content and uses them to "train our artificial intelligence/machine learning models." The policy even warns users not to include sensitive personal information in their interactions — an implicit admission that what you type isn't staying between you and the bot.

Independent reviewers put numbers on the concern. Character AI earns a C+ privacy grade, while of 15 AI companion apps reviewed, only 2-3 handled user data responsibly. A separate comparative analysis gives it an F safety score (22/100) against Replika's C (38/100).

Three details stand out for anyone skimming the fine print:

No opt-out from training. Character AI stores every message and uses all of it, with no setting to exclude your chats from model training and no encryption commitments.
No defined retention limit. The policy keeps data "for as long as necessary" — a phrase with no clock on it. Compare OpenAI, which processes deletions within 30 days.
A broad perpetual license. The Terms of Service grant a sweeping, perpetual license to exploit and commercialize user content.

The collection goes beyond text, too. Voice recordings, financial data, inferred demographics, and identifiers like Chat ID and Session ID feed what reviewers describe as a more detailed behavioral profile than competitors build.

Then there's the question nobody answers directly: does a human ever read your private chats? The policy discloses sharing with vendors "performing content moderation," and law enforcement via warrant or subpoena. But neither the official FAQ nor the policy states whether staff review individual conversations. That silence is worth noticing.

This is why privacy-first design matters at the platform level. FlirtAlly's approach — PG/Flirt by default, ID-verified opt-in for the 18+ "Spicy" side, hard-coded prohibited categories, and in-app reporting — treats safety as an explicit commitment rather than fine print. If you'd rather chat somewhere that puts honesty about monitoring up front, join the waitlist — be first in at flirtally.com.

You're Not Paranoid: Regulators and Experts Agree the Concern Is Real

If you've felt uneasy about what happens to your late-night conversations with an AI character, that instinct is backed by regulators, not paranoia. Character AI is currently facing an FTC probe, a Kentucky attorney general lawsuit, a Texas AG investigation, and a warning letter signed by attorneys general from 42 states. The platform has also been barred from allowing users under 18 to engage in open-ended chats at all.

That regulatory record exists because the concerns are structural, not hypothetical. Character AI's own privacy policy confirms it collects "chat communications" and uses them to train its AI and machine learning models — with no opt-out for users and no defined retention limit on how long those conversations are kept.

Third-party privacy assessments paint a similarly stark picture. CompanionWise's Safety Index gives Character AI an F grade, scoring just 22 out of 100 — compared to Replika's C grade at 38/100. A separate 2026 privacy review of 15 AI companion apps assigns Character AI a C+ for privacy, and found that only 2–3 of the 15 apps reviewed handled user data responsibly overall.

The reviewer at AI Companion Guides put the stakes bluntly: "The idea that some unknown company could be reading a 'anonymized' version of those conversations made my stomach turn." Character AI has improved some policies under regulatory pressure, but reviewers note that conversations are still used for model training, and data deletion remains only partial.

The opacity problem goes beyond any single policy document. Dr. Jennifer King of Stanford's Institute for Human-Centered AI told ABC News that once your data is used for training, "we don't really know... what those processes look like from the outside." Her advice is practical:

Avoid uploading sensitive files without redacting personal details first
Use temporary chat modes when a platform offers them
Opt out of data training wherever the option exists
Re-read privacy policies regularly, since terms change

Notably, neither Character AI's policy nor its support FAQ explicitly states whether human staff read individual private chats — a gap worth remembering before you share anything sensitive.

This regulatory and expert consensus is why the design choices behind newer platforms matter. FlirtAlly, for example, builds its honest safety approach into the product itself: a PG/Flirt default experience, ID-based age verification before any 18+ access, hard-coded prohibited content categories, media screening, and in-app reporting. Transparency about what's monitored — and why — is becoming the baseline users should expect from any companion app, not a premium feature.

How to Protect Yourself on Any AI Chat Platform

Protecting personal information on AI chat platforms is more crucial than ever. With platforms like Character AI actively collecting and using chat data for model training, users must take proactive steps to safeguard their privacy. According to Character AI's privacy policy, the platform collects chat communications to train its AI and machine learning models. This practice is not unique to Character AI; it is standard across the industry, including platforms like OpenAI, which also discloses monitoring and using chat content for various purposes.

To stay safe, consider the following expert-backed steps:

Never share sensitive personal information. Character AI explicitly warns that sensitive personal information is not required or necessary to use the services, and users should avoid including it in interactions.
Utilize available privacy controls. Platforms like OpenAI offer features such as temporary chats, disabling training, and deleting history. These controls help users manage their data and reduce the risk of personal information being used for training purposes.
Review privacy policies regularly. Privacy policies can change, and staying informed about how your data is being used is essential. For instance, Character AI's privacy policy was updated on July 1, 2026, and includes information on data collection and usage.
Opt out of training where available. Some platforms allow users to opt out of having their conversations used for training AI models. This is an important step in maintaining control over your data.
Leading AI-companion platforms like FlirtAlly prioritize honest safety, offering features such as ID-verified 18+ access and in-app reporting to ensure a secure user experience.

While platforms like Character AI are under regulatory scrutiny for their data practices, it is clear that user choices play a significant role in maintaining privacy. According to a recent study, 50% of American adults have used an AI tool, highlighting the need for awareness and proactive measures. Additionally, ZDNET emphasizes that users have control over their privacy through available settings and regular policy reviews. By taking these steps, users can better protect their personal information and enjoy a safer experience on AI chat platforms. Considering the unique privacy features of FlirtAlly, users can also explore these platforms for a more controlled and transparent experience.

A Different Approach: FlirtAlly's Honest Safety Model

Most AI companion apps have to be caught being careless before anyone talks about safety. A newer platform, FlirtAlly, is taking the opposite bet: putting its rules on the label before you ever send a message.

The contrast matters because the industry's baseline is grim. Independent reviewers graded Character AI's privacy practices a C+, and its safety score an F (22/100), citing no opt-out from training and no defined retention limit. Of 15 AI companion apps reviewed in a recent privacy guide, only 2-3 handled user data responsibly. FlirtAlly's answer isn't to promise that no one ever sees anything — it's to be explicit about what's restricted, how, and why.

That starts with a tiered experience. FlirtAlly's default mode is PG/Flirt — the standard experience for every member. The 18+ "Spicy" side is strictly opt-in, gated behind ID-based age verification and a qualifying plan (Flame and above at launch). Nobody stumbles into adult content by accident, and nobody has to guess what a subscription quietly unlocks.

Beyond age gating, FlirtAlly builds its limits into the product itself:

Hard-coded prohibited categories that no companion, creator, or plan can override
Media scene screening, so generated photos and videos are checked — every render is also identity-checked against the companion's canonical face
In-app reporting, so members can flag problems directly rather than hunting for a buried email address

This stance puts FlirtAlly outside the "unrestricted" arms race. Some platforms market themselves on having no filters at all — CrushOn.ai, for instance, earned an F privacy grade in the same review roundup. FlirtAlly treats that framing as a red flag rather than a selling point, on the theory that a platform honest about its boundaries is more likely to be honest about its data.

The regulatory backdrop suggests users are starting to agree. Character AI currently faces an FTC probe, a Kentucky AG lawsuit, and a 42-state attorneys general warning letter, and has been barred from allowing under-18 users into open-ended chats. As experts advise, the platforms worth trusting are the ones that tell you their rules up front.

FlirtAlly is in pre-launch, and registration opens through a waitlist. If you want a companion platform that treats honesty about safety as a feature — join the waitlist and be first in. You can also browse the companions' public profiles while you wait.

Can Character AI see your chats?