Privacy Policy
Last updated: 16 October 2026
This policy explains what AdventAI collects, why, who it is shared with, how long it is kept, and what you can do about it. It covers adventai.app, Bible World (bibleworld.ai), the AdventAI API and MCP server, and the installed apps.
In short
- We do not sell or rent your information, show ads, or build advertising or behavioural profiles.
- We do not use your content to train AI models. The AI providers that answer your chats do not train on it either; the one exception, the Translator, is named below.
- Outside providers receive only the content needed to do the task you asked for. We do not send them your name, email address or account id.
- Our staff read a conversation only when you send feedback on it, share it publicly, ask us for help with it, an automatic safety check flags it, or the law requires it. Every time is logged.
- You can download all your data and delete your account yourself, from your account page.
What we collect
Account details: your email address and name, and the sign-in method you use (email, Google or Apple, through Auth0). On Bible World, the username, display name, bio and profile picture you choose.
What you give us: chat messages, files you attach, voice input (as text, once transcribed), memories, projects, notes and documents, image and video prompts, reference images and uploads, shared links, saved library items, feedback and messages you send us.
What the service records: the IP address and browser your account was created from, device identifiers used for limits and abuse prevention (see Device identifiers), the cost and model of each AI request, messages sent per day, listening time, your library search history, API usage per key, and which policies you accepted and when.
Payments: Stripe processes payments. We never see your card number. We keep your plan, payment amounts and dates, and your Stripe customer id.
How we use it
- To provide the service: answer your questions, keep your conversations, files and projects, make images and videos, read text aloud.
- To apply plan limits and credits, and to stop abuse such as creating many accounts for free credits.
- To keep the service safe: automatic checks on chat messages and replies (see Who can see your content).
- To help you when you ask, and to send account emails (sign-in, receipts, invitations).
- To improve the service, using counts and costs (for example, how many answers used a tool or how much a model costs), not the content of your conversations.
- To meet legal obligations.
What we never do
- Sell, rent or trade personal information, or share it for advertising.
- Build profiles of you from what you write, how much you write or when you use the service, beyond the usage counts needed for plan limits.
- Use device identifiers or fingerprints for anything other than limits and abuse prevention, or to follow you across other websites.
- Sell or give aggregated or anonymised usage patterns to anyone. We may publish overall totals (for example, the number of conversations in a month), never data about a person.
- Use your content to train AI models, ours or anyone else’s (except the Translator, below).
Who can see your content
Your conversations, files and memories are private to your account. A small team of AdventAI administrators can open a conversation only for one of these reasons:
- You sent feedback on it (the feedback form says the team will read it).
- You shared it publicly.
- The automatic safety check flagged it.
- You asked us for help with it.
- The law requires it (see Legal requests).
Each opening is recorded with who opened it, when and why. Opening a conversation for any other reason is not possible in our admin tools.
Automatic safety check: an AI classifier (TypeSafe, listed below) checks each message, each reply as it is written, and text brought in from files and tools, for harmful content and attempts to misuse the assistant. It cannot be turned off. A flagged reply may be withdrawn, and the conversation is marked for review. No person reads a conversation because of this check unless it is flagged.
Feedback: when you send feedback on an answer, the question, the answer and your comment are emailed to the AdventAI team inbox so we can fix the problem.
Bible World: images you submit to the public gallery are reviewed by staff before they appear. Images made on a free account are released as public domain (CC0, see the Terms of Service); staff and an AI filter may review them and publish chosen ones in the public gallery under the AdventAI name, without your name or prompt.
Shared conversations are public: anyone with the link can read them, and they may appear in search engines.
Outside providers
We use the companies below to run the service. Each receives only the content needed for the task, under a contract that limits its use to providing that service. AI providers receive the text, files and images of the request. Files are passed as short-lived private links that the provider downloads. They do not receive your name, email address or account id.
OpenAI (GPT models, GPT Image)
- Used for
- Writes chat answers (the default model), suggests follow-up questions, makes images when you choose an OpenAI model, and answers decisions when TypeSafe is slow.
- Receives
- The conversation and the files and images in it; image prompts and reference images.
- Trains AI on it
- No. OpenAI does not train on API data.
- Keeps it
- Up to 30 days for abuse monitoring. We ask OpenAI not to store responses.
Google (Gemini API, Veo)
- Used for
- Writes answers in Big brain mode or when you choose a Gemini model; translates messages to and from English; describes and tags images; makes images and videos when you choose a Google model.
- Receives
- The conversation and its files; message text for translation; image and video prompts and reference images.
- Trains AI on it
- No. We use Google’s paid API, which is not used to train Google’s models.
- Keeps it
- Kept for a limited time for abuse detection. Generated videos are deleted from Google after 2 days.
Anthropic (Claude)
- Used for
- Names conversations, summarises long conversations and files, writes memories, plans searches, writes share descriptions and image titles; writes answers when you choose a Claude model.
- Receives
- The parts of the conversation, file or prompt needed for that task.
- Trains AI on it
- No. Anthropic does not train on API data.
- Keeps it
- Deleted within 30 days. If a request is flagged for breaking Anthropic’s usage policy, up to 2 years.
xAI (Grok, Grok Imagine)
- Used for
- Writes answers when you choose Grok; makes images and videos with Grok Imagine (the default video model).
- Receives
- The conversation; image and video prompts and source images.
- Trains AI on it
- No. xAI does not train on API data.
- Keeps it
- 30 days.
TypeSafe (Jev)
- Used for
- Makes the quick decisions in each chat turn: the safety check, which tools and sources to use, ranking passages, what to remember, which pictures fit.
- Receives
- The conversation and the passages being judged.
- Trains AI on it
- No. TypeSafe does not train on what it receives.
- Keeps it
- Only as long as needed to provide the service; TypeSafe states no fixed period.
SubQ (Snap)
- Used for
- In voice mode, decides whether you have finished speaking; detects the language of Translator text.
- Receives
- Short pieces of your voice transcript; the first 600 characters of Translator text.
- Trains AI on it
- No.
- Keeps it
- Up to 30 days.
Mistral AI (OCR)
- Used for
- Reads the text of scanned PDF pages you attach.
- Receives
- Up to 30 pages of the PDF, as images.
- Trains AI on it
- No. We have turned off Mistral’s use of our data for training.
- Keeps it
- Up to 30 days.
ElevenLabs
- Used for
- Turns your speech into text and reads answers and library text aloud.
- Receives
- Your microphone audio and the text to read. Audio goes straight from your browser to ElevenLabs, so ElevenLabs also sees your IP address; it never passes through our servers.
- Trains AI on it
- No. We have turned off ElevenLabs’ use of our data to improve its models.
- Keeps it
- ElevenLabs keeps request history for its own records; deleted data leaves its backups within 30 days.
Meta (Muse), and OpenRouter for the Translator
- Used for
- Translator page (staff and people given access only): Meta’s Muse Spark Contributor model writes the final translation, reached through OpenRouter. Muse Image makes images when you choose it.
- Receives
- The text you translate and your instructions; image prompts and reference images for Muse Image.
- Trains AI on it
- Translator text: yes, Meta may use it to train its models, so do not translate private or personal text. Muse Image: no. OpenRouter does not train on it.
- Keeps it
- Meta keeps requests as long as it needs them and states no fixed period. OpenRouter does not store the text.
BytePlus (Seedance)
- Used for
- Makes videos when you choose a Seedance model.
- Receives
- The video prompt and reference images.
- Trains AI on it
- No.
- Keeps it
- Generated videos are available for 24 hours. Requests stopped by its content filter are kept 180 days.
Amazon Web Services
- Used for
- Runs our servers (United States, Oregon) and stores your files (Australia, Sydney).
- Receives
- Everything the service stores.
- Trains AI on it
- No.
- Keeps it
- As long as we keep the data (see How long we keep data).
PlanetScale
- Used for
- Our database (on AWS in Oregon, United States).
- Receives
- Everything the service stores, except files.
- Trains AI on it
- No.
- Keeps it
- As long as we keep the data.
Auth0 (Okta)
- Used for
- Sign-in, including Google and Apple sign-in.
- Receives
- Your email address, name and sign-in details.
- Trains AI on it
- No.
- Keeps it
- Until your account is deleted.
Stripe
- Used for
- Payments.
- Receives
- Your email address, name, account id and the payment details you enter on Stripe’s page.
- Trains AI on it
- No.
- Keeps it
- Stripe keeps payment records as financial law requires.
Postmark
- Used for
- Sends account emails.
- Receives
- Your email address and the email’s content.
- Trains AI on it
- No.
- Keeps it
- 45 days. We do not track whether you open emails or click their links.
Mux
- Used for
- Hosts and plays generated videos, music videos and read-aloud audio.
- Receives
- The videos and audio; your IP address when you play them.
- Trains AI on it
- No.
- Keeps it
- As long as we keep the video. Its viewer analytics and cookies are turned off.
Bunny.net
- Used for
- Delivers public images: Bible World images, uploaded reference and profile images, shared-conversation cards.
- Receives
- The images; your IP address when you view them.
- Trains AI on it
- No.
- Keeps it
- Request logs 3 days, with IP addresses anonymised.
Adobe Fonts, Google Fonts, jsDelivr
- Used for
- Load the site’s fonts and the video player.
- Receives
- Your IP address and browser details, as with any web request.
- Trains AI on it
- No.
- Keeps it
- See their policies.
How long we keep data
- Your account and content: until you delete them or delete your account.
- A conversation you delete: removed from our database straight away, with its run records, any public link made from it and the memories learned from it. The record of what its AI requests cost stays, without their content.
- A file you delete: removed from storage straight away; the record of its name within a day.
- A Bible World image or video you delete: removed within 30 days, unless it was published publicly or chosen for the public gallery under the AdventAI name.
- Memories: removed straight away when you delete them on the Memory page; 30 days after you ask the assistant to forget one. Memories unused for six months are archived and no longer used.
- Your account, when you delete it: deletion starts at once and normally finishes within an hour. It removes your conversations, shared links, files, memories, projects, library data, Bible World images and characters, settings, sign-in and Stripe customer record. What stays: payment and invoice records (without your name or address), the cost of past AI requests (without their content), images already published under the AdventAI name, and an empty “deleted” marker where others replied to your comment.
- Device identifiers, IP addresses and abuse-prevention records: 12 months. The IP address and browser recorded when you accept a policy are kept as proof of acceptance while your account exists.
- Detailed records of how an answer was produced: 90 days.
- Server logs: 1 month.
- Records of AI requests (model, cost, time): kept for accounting, without their content.
- Payment records: as long as tax law requires (5 years in Australia).
- Backups: deleted data stays in encrypted database backups until they expire, at most 30 days.
There is no zero-retention mode: every conversation is saved until you delete it.
Device identifiers
To apply message limits, including for people who are not signed in, the AdventAI site computes a device identifier in your browser (FingerprintJS, open source, run on our site only) and notes whether the window is private. We store it with a one-way hash of your IP address and browser.
To stop people creating many accounts for free image credits, Bible World computes a one-way hash of your browser’s characteristics once per session when you are signed in, and stores it with your IP address and browser.
These identifiers are used only for limits and abuse prevention, are never shared, and are deleted after 12 months.
Your choices and rights
- Download your data: Account → Your data. You get a zip with every conversation (as Markdown and JSON), memories, projects, documents, your files (with download links valid for a day), library and listening data, Bible World and music data, API keys (without the keys themselves), billing records and the policies you accepted. Records kept only for abuse prevention and accounting (device identifiers, risk checks, AI costs) are not in the download; ask us for them by email.
- Delete your account: Account → Your data. Deletion is permanent. If you own an organization with other members or a subscription, close or hand it over first.
- Delete a conversation, file, memory or image at any time where it appears.
- Turn off memory on the Memory page.
- Correct your details on your account page, or ask us.
- Ask us what we hold about you, or object to a use of it, by email. We reply within 30 days.
You can complain to the Office of the Australian Information Commissioner (oaic.gov.au), or to the data protection authority where you live.
Legal requests
We give personal information to a government or court only when a valid legal order requires it. We check each request, push back on requests that are too broad, and hand over only what the order covers.
We tell you before we disclose, so you can respond, unless the law forbids it or someone’s life or safety is at risk. If we could not tell you first, we tell you once the restriction ends.
Security and where data is kept
Data is encrypted in transit and at rest. Our servers and database are in the United States (Oregon); uploaded files are stored in Australia (Sydney); AI providers process requests mainly in the United States. Access to production systems is limited to the administrators who run the service.
Children
The service is not meant for children under 13. If you are under 18, use it with a parent’s or guardian’s permission. If you believe a child under 13 has an account, tell us and we will delete it.
Changes to this policy
When we change this policy we update the date above. If a change affects how your information is used or shared, we ask you to accept the new version the next time you sign in.
Contact
AdventAI is run from New South Wales, Australia. Questions or requests about your information: hello@adventai.app