The short answer
Janitor AI is a real service with an official website and Android app, but that does not make every chat, public character or outside model equally safe. Treat it like any service that stores account information and accepts private conversation text: use a unique password, keep identifying details out of prompts and read the current policy before relying on it for sensitive conversations.
The official Google Play disclosure says the Android app may collect personal information, app activity and other data categories. It also says data is encrypted in transit and that users can request deletion. Those are meaningful signals, but they do not tell you how every connected model provider handles text or how every community character will behave.
Three kinds of safety people often mix together
Account safety
Whether sign-in, recovery and transport security protect access to your account. A unique password and secured email account still matter.
Conversation privacy
What chat text is stored, who processes it and how deletion works. This can become more complex when an outside model or API account is involved.
Content safety
What you may encounter in public characters and how prohibited material is moderated. A platform that permits adult roleplay still needs boundaries.
What the official Android disclosure actually says
Google Play's data-safety panel is the clearest official disclosure we could verify. It lists possible collection of personal information, app activity and other data types; says data is encrypted in transit; and says deletion can be requested. The same panel says no data is shared with third parties, based on the developer's declaration.
Store disclosures are summaries supplied by developers, not independent privacy audits. Use them as a starting point, then check the current privacy notice and any model-provider terms shown inside Janitor AI before entering information you would regret exposing.
A two-minute safety check before using Janitor AI
- 01
Secure the account
Use a password you do not reuse elsewhere and make sure the email account behind recovery is protected.
- 02
Identify the model path
Check whether the selected chat uses a Janitor-hosted option or an outside provider. A second provider can mean a second set of data rules.
- 03
Find deletion before you need it
Locate the account and data-deletion controls now. If the route is unclear, do not assume deleting a conversation removes every retained record.
- 04
Start without real-world details
Use a nickname and fictional context. Do not share addresses, employer information, financial details or private information about another person.
Can Janitor AI see your chats?
Assume the service and any model provider involved can technically process the text needed to produce a reply. “Private” usually means the conversation is not publicly posted, not that no system operator can ever access or retain it.
We did not find accessible official documentation during this review that justified a stronger promise about staff access, retention windows or training use. That uncertainty is exactly why sensitive real-world information should stay out of roleplay chats.
Public characters add a different risk
A community catalogue contains work from many creators. Character descriptions, greetings and example messages can be surprising, explicit or simply badly labelled. Check the character page before starting and leave if the scenario is not what it claimed to be.
Never treat a character's dialogue as professional medical, legal, financial or mental-health advice. It is generated roleplay, even when the character sounds confident or caring.
Extra caution with API keys
If you connect an outside model account, create a key for that purpose where the provider supports it, set spending controls and never paste a secret into a public character field. Revoke the key when you stop using the connection.
An API key can link usage and charges to another account. It also means both the chat interface and the model provider may be relevant to your privacy decision, so read both sets of terms.
How Sultrai approaches the same question
Sultrai does not ask users to connect a personal model API key. Private chats are not published as community posts, while publishing a character or selected media is a separate creator action. The service still has an Acceptable Use Policy and automated safeguards for prohibited categories.
That simpler setup removes one outside-account decision; it does not make sensible privacy habits optional. Use fictional details and review the privacy and deletion controls of any companion service you choose.