Are Character AI Chats Really Private?

Yes, Character AI chats are private from other users and character creators — but no, they are not private from Character AI the company.

That distinction matters more than it sounds.

Here's how privacy actually works on the platform:

• Other users. No one else on Character AI can see your conversations. Are character ai chats public? No — your chats are not visible to anyone else on the platform.
• Character creators. The person who built the character you're talking to cannot see your messages. Character AI's Help Center confirms this explicitly.
• Character AI staff. Staff can access your chats when flagged by automated moderation. Not routine surveillance, but not impossible either.
• Law enforcement. Character AI can share your data in response to subpoenas, court orders, and mandatory reporting obligations.
• Encryption. Your chats use TLS encryption in transit but are not end-to-end encrypted. Character AI can read messages stored on their servers.

Stanford HAI researcher Jennifer King put it bluntly when asked whether people should worry about privacy on AI chatbots: “Absolutely yes.”

Her research found that all six major U.S. AI companies use chat data by default for model training. Some retain it indefinitely. The privacy policy confirms Character AI collects chat communications, device info, IP addresses, and usage patterns — but does not specify how long that data is kept.

Can people see your chats on Character AI? No. Can people read your chats on Character AI? Also no.

But the company can, and that is the privacy question worth sitting with.

Does Character AI Save Your Chats?

Yes, Character AI saves your chats — along with significantly more data than most people expect.

The privacy policy (last updated August 27, 2025) lists what Character AI collects: personal identifiers, device information, IP addresses, usage patterns, cookies, voice data, and chat communications.

All of it sits on company servers. No specified retention timeline — they haven't committed to deleting your data after any particular period.

It goes further than storage. Character AI explicitly uses chat data to train and improve its AI models. If you're in the EEA or UK, you can opt out. Anywhere else? No documented opt-out mechanism exists.

Your conversations aren't just saved; they become training data that shapes how the AI talks to everyone.

Does Character AI collect data beyond chat text? Considerably — device identifiers, browser type, operating system, referral URLs, and behavioral usage data all appear in the policy. It's the kind of list that makes you wonder what they don't collect.

(The answer: not much.)

Does Character AI sell your data? The policy says they may share “aggregated or de-identified” information but doesn't define what “aggregated” means in practice.

Stanford HAI researchers found that major AI companies routinely merge chatbot interactions with other platform data. They keep it indefinitely.

Does Character AI record conversations? Yes. Does Character AI spy on you? “Spying” is a stretch — it is standard data collection, not covert surveillance. But everything you type is stored and used.

Can Character AI Creators See Your Chats?

No, Character AI creators cannot see your chats — the platform explicitly states that conversations are private between you and the character.

Character AI's Help Center confirms that creators have no access to conversations other people have with their characters.

Can creators see your messages in Character AI? No. What about Character AI developers — can they see your chats through the character they built? Also no.

Creators see download metrics and ratings, not messages.

One caveat worth knowing: group chats. If you're in a Character AI group chat or room, other participants can see everything you type.

But even then, the character's creator still cannot see the conversation unless they are an active participant in that specific room.

Does Character AI Read Your Chats?

Yes, Character AI can and does read your chats — but not routinely, and not all of them. Staff access is event-triggered, not constant surveillance.

The question “does Character AI read your chats” implies a human scrolling through your messages. What actually happens is a layered system where automated tools scan everything, and humans only get involved when something gets flagged.

Every message passes through automated content moderation scanning for safety violations — self-harm language, CSAM indicators, violent threats, and filter triggers. Most of the time, no human sees it.

When the system flags something, it escalates to the trust and safety team. They may review flagged messages and your broader recent chat history for context.

How often does Character AI staff read your chats? No public data exists on volume, but the process is event-based, not routine.

Can Character AI staff see chats? Yes, under those conditions. Can Character AI employees, mods, or devs see your chats?

Same answer — moderation is primarily automated, but human staff on the trust and safety team can access flagged content. Engineers have server-side access to data too.

Here's something we noticed firsthand: when we tested Character AI in April 2026, the content filters triggered on emotional roleplay that had nothing to do with explicit content. A scene involving grief and comfort — nothing sexual, nothing violent — got flagged and interrupted mid-conversation.

That aggressive filtering means more conversations are likely being escalated to human review than you'd expect. If the filter is the dealbreaker, we covered why it's tightening in our guide to does character ai allow nsfw.

The security track record adds context. In December 2024, Character AI had two separate data exposure incidents.

In one, parts of affected people's account pages — usernames, bios, personas, and chats — were publicly visible for roughly 10 minutes. The company acknowledged “this incident was a failure of that trust.”

In the other, people reported being logged into strangers' accounts entirely. Full chat histories. Identifying information. One person called it “one hell of a mess.”

Does Character AI monitor chats? Yes — continuously, and with only a 6-character minimum password and no 2FA protecting accounts.

That's below industry standards by any reasonable measure.

What About Group Chats and Rooms?

Group chats and rooms operate under different rules. Every participant can see everything you type.

Are Character AI group chats private from outsiders? Yes. From other participants? No.

Can Character AI Report You to the Police?

Character AI can share your data with law enforcement — and in some cases, they're legally required to.

Three scenarios apply.

Subpoenas and court orders. Like any U.S. technology company, Character AI must comply with valid legal demands for data.

The Electronic Frontier Foundation noted in December 2025 that “law enforcement is already demanding user data from AI chatbot companies, and it will only increase.” They called chat logs “digital repositories of our most sensitive and revealing information”.

Mandatory reporting. Under 18 U.S.C. Section 2258A, providers who obtain “actual knowledge” of apparent child sexual abuse material must report to NCMEC's CyberTipline.

This is federal law, not company policy — Character AI has no discretion here.

Voluntary disclosure. The company can share data when they believe there's imminent risk of serious harm.

The threshold for what qualifies is undefined in the privacy policy, which is itself a problem.

Legal pressure is already building. Garcia v. Character Technologies (October 2024) was the first wrongful death lawsuit against an AI chatbot company.

The Kentucky AG filed the first state data privacy lawsuit against C.AI in January 2026. The FTC issued 6(b) orders in September 2025 to seven companies including Character Technologies. The focus: how AI companion platforms monitor impacts on teens.

Is Character AI Encrypted?

Character AI uses TLS encryption for data in transit, but your chats are not end-to-end encrypted — which means the company can access them on their servers.

The difference matters enormously.

TLS protects data while it travels between your device and Character AI's servers — think of it as an armored truck moving your messages. Nobody can intercept them on the road.

But once the truck arrives at the warehouse, the contents are unpacked and stored in readable form. Anyone with warehouse access can look at them.

End-to-end encryption locks the warehouse too. When WhatsApp says your messages are encrypted, they mean even WhatsApp cannot read them.

Character AI uses the armored truck. It does not lock the warehouse.

And honestly, that's the whole privacy question right there.

Training AI models on chat data requires the company to read that data. End-to-end encryption would make that impossible.

That is the architectural trade-off Character AI has made — deliberate, not an oversight.

In February 2026, security researchers exposed 53MB of source code from Persona — the third-party provider Character AI uses for age verification. The leak revealed practices like storing facial biometrics for up to three years.

How strong is a security chain when one of the links is already leaking source code?

What Happens When You Delete a Character AI Chat?

When you delete a Character AI chat, it disappears from your interface — but that does not mean the data is gone from Character AI's servers.

The privacy policy doesn't specify what happens server-side after deletion. No retention timeline, no documented deletion schedule, and no confirmation that data is actually purged.

Can Character AI staff see deleted messages? We do not know, because the policy simply doesn't address it — which, if you're feeling charitable, is an oversight; if you're not, it's a design choice.

The training data question makes this murkier. Character AI uses chat data to train its models — if your conversations entered the training pipeline before deletion, that influence does not get reversed.

You can delete the message, but you can't un-teach the model.

Does Character AI keep deleted chats? Possibly. Even if the raw text is purged, the derivative training data remains.

EU and UK residents can submit a GDPR Article 17 “right to erasure” request. California residents have a similar right under CCPA Section 1798.105. Contact privacy@character.ai to start the process.

But “erasure” when data has already trained a model is an unresolved legal question — one that regulators across every jurisdiction are still figuring out.

What About Teen Accounts and Parental Controls?

Sixty-four percent of U.S. teens have used AI chatbots, with 9% specifically using Character AI (Pew Research, 2025).

Teen accounts now operate under significantly stricter privacy and safety rules than adult accounts — but parents still cannot read their teen's actual chat content.

The Parental Insights tool, launched in March 2025, gives parents weekly reports on usage time and top characters. Chat content is explicitly excluded.

By November 2025, Character AI eliminated open-ended chatbot access for all people under 18 entirely. They replaced it with curated creative features — the most aggressive teen restriction of any major AI companion platform.

The pressure behind these changes is considerable.

The Garcia v. Character Technologies wrongful death lawsuit, the Kentucky AG's data privacy suit, and the FTC's September 2025 inquiry all involved teen safety.

CEO Karandeep Anand acknowledged in February 2026 that “the longitudinal impact of chatbot interaction could be unhealthy, or is not fully understood.”

And then there's the age verification system itself. A behavioral analysis model combined with third-party provider Persona flags people it thinks are under 18. Those flagged must submit selfie photographs and potentially government ID.

The February 2026 Persona source code leak revealed biometric data stored for up to three years. The verification process ends up creating privacy risks for the very young people it is designed to protect.

Does adding more data collection to protect teens from data collection actually make them safer?

Which AI Chat Platforms Are Actually Private?

The most private AI chat platform we've found is ourdream.ai — it's the only major option that uses end-to-end encryption, meaning not even the company can read your conversations.

The encryption question is the one that matters most — everything else is policy, and policy can change overnight.

• ourdream.ai. End-to-end encrypted, so your conversations stay between you and your companion. The only platform on this list where the company literally cannot read your messages.
• Character AI. TLS only. Staff can read chats when flagged. Chat data used for training with no opt-out outside EEA/UK. Aggressive content filtering.
• ChatGPT. Different use case entirely, more productivity tool than companion. Decent opt-out options, but still TLS-only.
• Replika. Decent privacy practices but limited creative freedom. Fine for casual use, restrictive for anything else.
• Chai. Lightweight but less transparent on data practices than any other platform here; documentation reads more like a placeholder than an actual privacy commitment.

ourdream.ai also has no NSFW content filters — so there's no risk of conversations being interrupted by a misfiring filter. We saw this firsthand with Character AI, where filters misfired on emotional roleplay that wasn't even remotely explicit.

For the full breakdown of unrestricted chat on ourdream, see our guide to ai sex chat and our roundup of the best nsfw ai chat platforms in 2026.

The platform is browser-only with no mobile app yet, which is a real limitation if you want a mobile experience.

But on the privacy question? The architecture is fundamentally different from everything else on this list.

How Can You Protect Your Privacy on Character AI?

You can protect your privacy on Character AI by controlling what you share, how you connect, and what rights you exercise — here are the steps that actually matter.

• Don't share real personal information in chats. No real name, location, school, or phone number. Character AI stores everything, and that data is accessible to staff and potentially law enforcement.
• Use a unique email address. Create a throwaway email for your C.AI account. The December 2024 breaches showed this isn't a theoretical risk — people's account data was exposed publicly.
• Use a VPN. It masks your IP address and approximate location from Character AI's data collection. Won't make you anonymous, but strips one identifying layer.
• Review and delete chat history regularly. Deletion may not remove server-side data, but it reduces accessible content tied to your account.
• Exercise your data rights. EU/UK residents can submit GDPR deletion requests; Californians can invoke CCPA. Contact privacy@character.ai.
• Avoid school or work networks. Administrators can see you visited character.ai, even if they can't read the chats themselves.
• Consider alternatives for sensitive conversations. End-to-end encrypted platforms are the only architecture that guarantees the company cannot read your messages. That's not a marketing line — it is a technical fact.

All of which raises a bigger question.

What Does It Mean When Your Private Thoughts Aren't Actually Private?

It means something different than it used to — and we haven't fully reckoned with that yet.

People use AI chatbots for grief processing, emotional support, sexual exploration, creative storytelling — confessions they wouldn't make to another human being. These platforms become repositories of vulnerability.

Every major AI company in the space collects that data by default. They train on it and retain it for periods they do not disclose.

The EFF was not exaggerating when it called AI chat logs “digital repositories of our most sensitive and revealing information.”

But the platforms provide genuine value. For some people, an AI chatbot is the only judgment-free space they have. Dismissing that would be cheap — and it would also be wrong.

The tension is real: the same intimacy that makes these tools meaningful is what makes the data so valuable to the companies that collect it. And so dangerous when it falls into hands it was never meant for.

Whether privacy regulation will catch up to the reality of AI chat data remains genuinely unclear. Deletion cannot undo what a model has already learned.

We're in uncharted territory, and the people making the rules haven't caught up to the people building the products.

FAQ

Where to Start

Your conversations on Character AI exist in a space that is neither fully private nor fully public. Private from other people on the platform — yes. Private from the company, its automated systems, its trust and safety team, or any legal authority with the right paperwork? No.

That is not unique to Character AI. It's the default posture of the entire AI chatbot industry.

A few platforms are starting to build differently, with encryption that means the company itself cannot read your messages. They remain the exception.

If end-to-end encryption is the line you want between yourself and an AI company — start with ourdream.ai. No training on your chats. No staff-accessible servers. No filter walls. Just the conversation you actually wanted to have.