Janitor AI is safe enough for casual roleplay if you treat every chat as something the company can store and, in some cases, review. It doesn't sell your data or show your private chats to other users, but your conversations aren't end-to-end encrypted. If you plug in a proxy, the company behind that proxy sees everything you write too.
This guide covers who can actually see your chats, what Janitor AI keeps, where the real risks sit, and how to delete your account. It draws on Janitor AI's privacy policy (last updated 17 September 2026), its safety page and its help center.
- Other users can't see your chats unless you publish one. Public chats arrived on 31 March 2026, and a published chat is readable by anyone.
- Janitor AI stores your chats on its servers. Its policy allows moderation and abuse investigations, and it promises encryption in transit, not end-to-end encryption.
- Proxies are the biggest real risk. With your own API key, every message also goes to that provider (OpenRouter, DeepSeek and others) under its own data rules.
- No Janitor AI breach is on record. Have I Been Pwned has no entry for it as of October 2026, and the company says it doesn't sell personal data.
- Sign up with an email alias and keep real names, photos and workplace details out of chats. That limits the damage if anything leaks.
What Is Janitor AI and How Does It Work?
Janitor AI is a character chat platform: you pick or build a character, and a language model plays it in a long-running conversation. It launched in 2023 and is known for looser content rules than Character.AI, which is why it attracts roleplay writers and an adult audience. Its terms set a minimum age of 18.
The privacy picture depends on which model answers you. There are three routes:
- JanitorLLM (JLLM): Janitor AI's own free model. Your messages stay with Janitor AI and the infrastructure providers it uses.
- janitor+ models: the paid tier adds third-party models that Janitor AI routes for you.
- Your own proxy and API key: you connect an outside provider and paste in a key. Janitor's beginner's guide to proxies calls a proxy “simply a bridge” to other AI models.
Every message you send includes the character card, your persona and recent chat history, because the model needs that context to reply. So whoever runs the model receives the whole conversation, not just your last line.
Can Anyone See Your Janitor AI Chats?
This is what most people are asking, and the answer differs by who's looking.
| Who | Private chats | Published chats |
|---|---|---|
| Other users | No | Yes, anyone can read, comment and fork |
| Bot creators | No. They only see the public chat count on their character | Yes, and they can hide published chats on their bot |
| Janitor AI staff and moderators | Possible, e.g. for reported content or abuse investigations | Yes, checked against the content policy before going live |
| Your proxy or model provider | Yes, every message you send through it | Yes |
Other users
Private chats stay private by default. The exception is public chats, which Janitor AI announced on 31 March 2026. Publishing saves a snapshot of the chat that anyone can read, react to, comment on or fork into their own private copy. If you shared your persona details, readers see those too. Unpublishing takes one click, but anyone who already copied or screenshotted the chat keeps their copy.
Bot creators
No. Janitor AI's privacy policy lists every situation in which it discloses your information (section 6): service providers working on its behalf, a merger or sale of the company, legal requests, and cases where you give explicit consent. Other users and bot creators aren't on that list. Publishing a chat is how you give that consent for one conversation.
The features match the policy. Every creator control in the public chats announcement applies to published chats only: a creator can switch publishing off for their character, hide published chats from its page and limit comments. There's no creator tool for opening a private conversation, and the only chat data a creator sees is the public chat count on their character. So the person who made “your” character doesn't get a window into your roleplay.
Janitor AI staff and moderators
This is where “private” has limits. Janitor AI's privacy policy lists chat conversations among the user content it collects. It says support and security staff may access account information to handle requests and investigate abuse. The safety page describes AI screening plus human moderators who review reported content around the clock.
In practice, human review is triggered by reports and automated flags, not someone browsing your chats for fun. But the technical ability exists, because the chats aren't end-to-end encrypted. Write as if a moderator could one day read a flagged conversation.
What Data Does Janitor AI Collect and Keep?
According to the privacy policy dated 17 September 2026, Janitor AI collects:
- Account details: email, username, password, profile details and, if you use Google sign-in, your Google account ID, name and profile picture.
- What you create: characters, personas and chat messages.
- Payment data for janitor+: billing address and transaction history. Card details go to the payment processor (Stripe) and aren't stored on Janitor AI's servers.
- Device and usage data: IP address, device identifiers, approximate location from your IP, server logs and cookies.
Four details matter most for privacy:
- No sale of data. The policy states that Janitor AI doesn't sell personal data under California's CCPA/CPRA, and the safety page says it won't share private conversations for advertising.
- Model improvement. The policy mentions “research and development to improve our AI systems” but doesn't say whether your chats train JLLM. Assume they could.
- No fixed retention period. Data is kept “as long as necessary”, and some records (such as your email and login metadata) may outlive account deletion for security, fraud and legal reasons.
- Encryption in transit only. The policy mentions HTTPS and encryption but makes no end-to-end encryption claim, and it notes that no method is completely secure.
Proxies and API Keys: The Biggest Real Risk
Most Janitor AI privacy horror stories trace back to proxies, not to Janitor AI itself. When you route chats through your own API key, two things change.
First, a second company holds your conversations. OpenRouter, DeepSeek or whichever provider you connect receives every prompt with its full context, and that provider's retention and training rules apply, not Janitor AI's. Some providers log prompts or use them for training unless you opt out in their settings.
Second, community-run “free” reverse proxies add an operator you know nothing about. Whoever runs one can log every message that passes through and see the key it uses. If someone asks you to paste your paid API key into a proxy they host, they can also spend your credits.
Safer habits if you use a proxy:
- Use the provider directly with your own key, rather than a shared community proxy.
- Check the provider's data settings. Turn off prompt logging or training where the provider allows it.
- Set a spending limit on the API key, and rotate it if you ever posted it in a Discord server or screenshot.
- Never share your key. Janitor's own guide treats it as “your unique password”, and it is.
Age Verification: What Janitor AI Checks and What It Keeps
Janitor AI is adults-only, and in most countries age is still self-declared at sign-up. In three countries it's now verified. According to Janitor AI's help center, users connecting from Australia and Brazil must complete a one-time check from 24 April 2026, and users in the UK from 15 June 2026. The requirement follows local laws: Australia's Online Safety Act 2021, Brazil's Digital ECA and the UK's Online Safety Act.
The check is run by k-ID, the same vendor Discord and Snapchat use. You can choose a selfie video processed on your device, a photo of a government ID, or an AgeKey reused from an earlier Discord or Snapchat check. Janitor AI says it receives only a pass or fail, that ID photos are deleted after confirmation, and that verification data isn't linked to your account. Detection is based on your IP address.
If you're in one of these countries, the privacy trade-off is real but limited. The selfie route keeps your face on your device, so pick it over the ID scan if you can.
Has Janitor AI Been Hacked or Leaked?
No breach of Janitor AI has been confirmed. As of 20 October 2026, Have I Been Pwned, the largest public index of data breaches, has no entry for the janitorai.com domain, and Janitor AI hasn't announced one. Searches for “Janitor AI leak” usually turn up three other things:
- Leaked API keys that users posted publicly or pasted into untrusted proxies. That's the key owner's exposure, not a Janitor AI breach.
- Characters copied and re-uploaded elsewhere after creators deleted them or staff removed them.
- Service problems such as lost chat memory or outages, which affect your data's availability but don't expose it.
No breach so far doesn't mean no breach ever, and other AI chat services show what one looks like. In September 2024, data was stolen from the AI companion site Muah.AI, including users' prompts; Have I Been Pwned lists about 1.9 million affected email addresses. In February 2026, Malwarebytes reported that the chatbot app Chat & Ask AI had left a Firebase database open, exposing about 300 million messages tied to 25 million users.
Any service that stores millions of intimate conversations is a target, and Janitor AI's policy is candid that no security method is perfect. Plan for the scenario where your chat history is exposed, and make sure it can't be tied to your real identity. Leaked email addresses are also the raw material for targeted scams, as covered in our guide to AI-powered phishing.
How to Use Janitor AI More Safely
The goal is simple: if your chats ever leaked, nobody should be able to work out they're yours.
- Sign up with an alias, not your main email. Your inbox ties together your bank, work and social accounts. A separate address keeps your Janitor AI account out of that web, and you can retire it at any time.
- Skip Google sign-in if you want separation. It's convenient, but it links the account to your Google identity and shares your name and profile picture, one more thread in the profile Google already builds about you.
- Keep identifying details out of chats and personas. No real names, school or workplace, city, photos of yourself or other real people. Each detail adds to your digital footprint.
- Use a unique password stored in a password manager, and turn on two-factor authentication on the email account you sign up with.
- Think before publishing a chat. Strip out personal details and your persona first. A published chat can be forked and copied.
- Be careful with proxies (see above). Use the provider directly and keep your key private.
- Delete chats you'd regret rather than letting them pile up for years.
Where Atomic Mail fits
Step one is the one most people skip, because creating a new email account feels like a chore. With Atomic Mail aliases, you create a separate address in a few seconds, use it only for Janitor AI and similar apps, and still receive the login and verification emails in one encrypted inbox. If that address starts getting spam or turns up in a leak, you delete the alias without changing your main email. Sign-up doesn't ask for a phone number, so the inbox isn't tied to your SIM either.
An alias won't protect what you type into a chat. It does keep a leak of your chat account from leading straight back to the email you use for everything else.
Create a free Atomic Mail account and make a separate alias for every chatbot you sign up for.
No phone number needed. Delete an alias whenever you're done with it.
How to Delete Your Janitor AI Account and Data
Janitor AI's privacy policy describes the process:
- Save anything you want to keep. Copy out characters or chats first, because permanent deletion can't be undone.
- Cancel janitor+ first. If you subscribed through the App Store or Google Play, cancel there, since deleting your account doesn't stop a store subscription.
- Go to Settings → Security → Delete Account. The account is deactivated first and can be restored during a grace period. After that, deletion is permanent.
- Request removal of retained data. Some records, such as your email and login data, may be kept for security and legal reasons. You can email the privacy contact in the policy to ask for those to be deleted, using your GDPR or CCPA rights where they apply.
- Clean up the other end. If you used a proxy, deleting your Janitor AI account doesn't delete what that provider stored. Delete or revoke your API key and check the provider's own data-deletion options.
If you signed up with an alias, also delete the alias. That cuts off any future email to that address.
Is Janitor AI Safe for Kids and Teens?
No. Janitor AI is built for adults, allows mature content and sets a minimum age of 18. Outside Australia, Brazil and the UK, that age is only self-declared, so a teenager can get in by ticking a box. The safety page says suspected minors may be suspended or banned, and content involving minors is banned outright.
For parents, the practical steps are the usual ones: device-level parental controls, blocking adult sites at the router or DNS level, and an honest conversation about why AI companions aimed at adults aren't built with teen safety in mind. If your teen wants an AI chat app, look for one with a real teen mode and published child-safety policies.
Safer Alternatives If You Want Full Control
If your roleplay involves things you'd never want on someone else's server, the only setup that keeps chats entirely yours is running a model locally. Front-ends such as SillyTavern can connect to a model running on your own computer (through tools like LM Studio, Ollama or KoboldCpp), so nothing leaves your machine. The trade-offs are real: you need a capable GPU or plenty of RAM, setup takes an evening, and local models are usually weaker than the large cloud ones.
Moving to another cloud chat site changes whose servers hold your data, not the basic risk. Character.AI, Chai and similar apps also store conversations, so the same rules apply: alias email, no identifying details, delete what you don't need. If you also run AI agents with access to your files or accounts, the risks are different again; see our OpenClaw safety guide.
So, Is Janitor AI Safe? A Quick Verdict
If you use JLLM, sign up with a separate email and keep real-life details out of your chats, Janitor AI's risks are about the same as any other cloud chat service. If you route chats through a free community proxy or publish chats containing personal details, you're the one making the data public. Fix those two habits and you've removed most of the risk.
FAQ
Is Janitor AI safe to use?
For most adults, yes, with sensible precautions. Janitor AI doesn't sell personal data and keeps chats private from other users, but it stores them without end-to-end encryption, so don't share anything that identifies you.
Can Janitor AI see your chats?
Janitor AI stores your chats, and its staff and moderators can review content, mainly when it's reported or flagged for abuse. Other users can't see private chats.
Can bot creators see my chats on Janitor AI?
No. Janitor AI's privacy policy doesn't list other users or creators among the parties it shares your data with, and creators have no tool to open private chats. They only see a chat if you publish it.
Are Janitor AI chats private?
They're private from other users by default. They're not private from Janitor AI itself, or from any proxy provider you route them through.
Does Janitor AI sell your data?
Its privacy policy says it doesn't sell personal data under California law, and its safety page says it won't share private conversations for advertising.
Has Janitor AI been hacked?
No breach has been confirmed: Have I Been Pwned has no Janitor AI entry as of October 2026. Most “leaks” people mention are API keys users exposed themselves or characters re-uploaded elsewhere.
Is Janitor AI encrypted?
Data is encrypted in transit (HTTPS), but chats aren't end-to-end encrypted, so Janitor AI can technically access them.
Do I need to verify my age on Janitor AI?
Only if you connect from Australia, Brazil or the UK, where a one-time k-ID check is required. Everywhere else you confirm you're 18 or older at sign-up.
Is it safe to use a proxy with Janitor AI?
It's safe if you use a reputable provider directly with your own key and a spending limit. Free community proxies are risky, because their operators can log your chats and see your key.
How do I delete my Janitor AI account?
Go to Settings → Security → Delete Account. The account can be restored during a short grace period, then it's deleted permanently. Cancel any app-store subscription separately.
