Privacy Policy
Last updated: 16 October 2026
This page explains what Chatbots keeps, why we keep it, and who else receives it.
What we collect
- Your Google profile basics: name, email address, profile picture link and Google account ID, received when you sign in with Google. If Google gives no name, we use the part of your email before the @. We refresh these details each time you sign in with Google.
- Username and password: only if you make an account without Google. The password is stored as a salted hash, never as you typed it.
- Your last sign-in: the date and time you last signed in.
- Account settings: whether your account is an admin or is banned, whether Claude is turned on for it, and any Claude limit the site owner has set for it.
- Usage counts: how many messages you send and how many files you upload per day (UTC), to enforce the daily allowances.
- Your saved chats: your messages and the replies, and what we read from the files you attach: each file's name and kind, a PDF's page count and text, transcripts of audio, descriptions of images and a small (256 pixel) thumbnail of each image. A chat whose first message has no text is titled with the name of its first file. Each account keeps up to 100 chats.
- Claude billing: for each Claude reply, the model used, what the reply cost us, the amount billed to you, and the monthly invoice it is part of. When billing is on, we also keep invoice records and their status, and a Stripe customer ID that links your account to Stripe. Card and bank details are entered on Stripe's own page, and we never receive them.
- Cookies: a session cookie that keeps you signed in for up to 30 days, and a short-lived cookie used only while you sign in with Google.
- Network address: used in memory only, to slow down sign-ups, guest visits and sign-in attempts. We don't save it in our database.
What we don't keep
We do not keep the files you upload. From a PDF or an audio recording, we keep only the text we read, plus the file name, the file kind and, for a PDF, the page count. From an image, we keep a short description, a small thumbnail, the file name and the image type. Full-size images are never kept.
We do not keep token counts for Claude replies. We never see card or bank numbers. We do not save your network address in our database. Guest messages are not saved.
How long we keep chats
A saved chat stays until you delete it. Once you have more than 100 chats, the chats that have gone longest without a new message are removed, along with their messages. Opening a chat does not count.
How long we keep other data
Account records, usage counts, Claude billing records (one per Claude reply) and invoice records are kept. Nothing in the app removes them automatically.
A sign-in lasts 30 days. It renews each time you use the site when less than 15 days are left. Old sign-in records are not removed automatically.
Chats from the earlier version of the site
Before sign-in existed, the site saved chats in our database, linked to the browser you used rather than to an account. When sign-in was added, those chats were moved into separate tables. The site never reads them, and they are not linked to any account. Nothing removes them automatically, and the site has no way to delete them.
Guest visits
You can try the site without signing in. A guest visit creates a record with no personal name, email or Google account. The record also counts the messages sent. Each network address can make five guest visits a day. Guests have no daily message limit.
Messages from a guest go to the model you pick, but they are not saved. Guests can't attach files or use Claude. Guest records are not removed automatically.
Who sees your messages
When you send a message, it goes to the AI provider that runs the model you picked, so the provider can write a reply. The providers are Groq, Google (Gemini), OpenRouter, Anthropic (Claude), and Ollama if the site owner has turned it on.
Along with your message, the provider receives the earlier messages in the same chat (up to 200) and the text we read from your files. Google and Anthropic models also receive the image itself. Anthropic does not receive HEIC or HEIF images. We never send your name or email to an AI provider. Each provider's own privacy policy applies. The Service providers section says what each one receives.
Messages from signed-in users are also checked by OpenAI's moderation service, which sends back whether a message breaks our terms. The message text is sent to OpenAI for that check. If a message is flagged, we keep a short excerpt of it (up to 300 characters) for a site administrator to review, and delete it after 30 days. Guests are not checked. OpenAI's privacy policy and terms of use apply to that check.
Who can see your data
The site owner can see account details, usage counts and Claude usage in the admin panel. The owner can also change an account's role and limits, turn Claude on or off for it, and ban it. The admin panel does not show anyone's chats.
Saved chats are stored as readable text in the database. The site owner, who runs that database, could read them. We do not use your chats to train AI models. We do not sell your information, and we do not use it for advertising.
Hosting companies and the other providers listed below can see data as it passes through their systems. When billing is on, Stripe receives billing details, as described below.
The site loads no analytics or tracking scripts.
Service providers
This section lists each outside company that receives data when you use Chatbots. Each entry says what the company receives, why, and whether it is always on. Links open in a new tab.
Other sign-in options (Discord, Microsoft, Facebook, X)
- What they receive: When you choose one of these, your browser goes to that company's sign-in page, and the company sends us your account ID and the profile details it shares: name, profile picture link, and, where it shares them. We do not keep the email address from any of these companies, so these accounts can never be given admin rights. X does not share an email at all. Accounts from these companies are kept separate from Google accounts, so the same person may have more than one account. We never send chats or files to these companies.
- Why: So you can sign in with that account. We use the company's account ID to tell accounts apart.
- On or off: Each option is on only when the site owner has set it up.
- Their policies: Discord, Microsoft, Facebook and X.
Google sign-in
- What it receives: When you sign in, your browser goes to Google's sign-in page. Google receives our app's ID, the page it sends you back to, and some random values that keep sign-in safe. Google also sees your network address and browser details, as it does on any sign-in page. Google then sends us your Google account ID, your email address, whether Google has verified that address, your name and your profile picture link. We accept a sign-in only when Google has verified the address. Our server also calls Google to finish the sign-in. We do not send chats or files to Google.
- Why: So you can sign in with your Google account. We use the account ID and email to tell accounts apart, and we keep the name and photo link to show on the site.
- On or off: Always on.
- Company policies: Google privacy policy · Google terms · Google API Services User Data Policy · Google APIs Terms of Service · Google OAuth 2.0 policies
Google profile photos
- What it receives: Your browser loads your profile photo from Google's image server, so Google sees your network address and browser details. Your browser sends no referring page address with that request. Our server does not download the photo. It only stores the link and passes it to your browser.
- Why: To show your photo in the site header and in the admin user list.
- On or off: Only for Google sign-in accounts that have a photo. No setting on the site turns it off.
- Company policies: Google privacy policy · Google terms
Stripe (card and bank payments)
- What it receives: Only when billing is on. When you set up a card or US bank account, Stripe receives your email address, unless you signed up with a username and password, because those accounts have no email. Stripe also receives a customer ID, and an internal account ID. For each monthly invoice, Stripe receives the amount, the month, and a short description such as "Claude usage for 2026-10 (12 replies)". Stripe hosts the card and bank form, so card and bank numbers go to Stripe and not to us. Chat text and files are not sent to Stripe. Stripe does not receive your name or your Google account ID. Stripe sends us payment updates, such as paid and failed invoices and refunds.
- Why: To take payment for Claude use. Stripe saves the card or bank account, sends monthly invoices (only when the balance reaches $5), records payment status, and handles refunds.
- On or off: Off unless the site owner sets up Stripe and turns on payments. While payments are off, no payment is taken.
- Stripe's own use: Stripe also uses some of this data for its own purposes, such as fraud checks and improving its products. Stripe's privacy policy says it may use personal data to train its AI models. We have not confirmed whether that covers the invoice and account records we send.
- Company policies: Stripe privacy policy · Stripe Services Agreement · Stripe data processing agreement · Stripe service terms (Billing)
Groq (chat models)
- What it receives: For each chat request: the earlier messages in the chat (up to 200), the site owner's instructions for the model, the text we read from attached files, the model name, and any settings for how the model writes replies. Groq receives text only. It does not receive images, your name, your email or your account ID.
- Why: To run the free Groq models that the site owner has turned on.
- On or off: Optional. Used only when the site owner has saved a Groq key.
- Use and keeping: Groq's services agreement says Groq may not use your inputs or outputs to train models, unless the customer allows it. By default, Groq does not keep data from inference requests. It may keep inputs and outputs for up to 30 days to troubleshoot and check for abuse. The customer can turn on Zero Data Retention in the Groq console, which stops that. Whether it is on for the key this site uses depends on the site owner's Groq account.
- Where: Groq stores customer data in the United States and may process it in the United States and other countries.
- Plan: Groq's agreement treats some free-of-charge and beta services differently, and its data addendum may not cover them. The Groq plan for the site's key is set in the Groq console.
- Company policies: Groq privacy policy · Groq services agreement · Groq data processing addendum
Groq Whisper (audio transcription)
- What it receives: The audio file you upload, such as a voice note, with its file name, and the name of the transcription model. Groq does not receive your name, email or account ID.
- Why: To turn the audio into text. The transcript is attached to your chat as text.
- On or off: Optional. Used only when the site owner has saved a Groq key. If Groq is not set up, fails, or returns an empty transcript, the audio goes to Google Gemini instead. If neither key is set, audio uploads are refused.
- Use and keeping: The Groq rules above are assumed to cover audio. Groq's speech page does not say so directly, and it gives no separate keeping period for audio.
- Company policies: Groq privacy policy · Groq services agreement · Groq data processing addendum
OpenRouter (chat models)
- What it receives: The same text-only data as Groq: the earlier messages in the chat, the site owner's instructions, the text from attached files, the model name and any settings. It also receives the app's name, Chatbots. It does not receive images, your name or your email.
- Why: To run the chat models served through OpenRouter that the site owner has turned on.
- On or off: Optional. Used only when the site owner has saved an OpenRouter key.
- Use and keeping: OpenRouter says it does not use your inputs or outputs to train models. Prompt logging is off by default. Each model provider that handles a request has its own rules, and some may use inputs to train or improve their models. Which provider handles a request depends on routing and on the site owner's privacy settings in OpenRouter. If the owner turns on prompt logging, OpenRouter's terms give it a license that never ends and cannot be withdrawn. It covers use for OpenRouter's own business, including licensing or selling the content in anonymized form, once it is no longer linked to you.
- Where: OpenRouter may store data in the United States or in other countries.
- Plan: The owner's OpenRouter plan and privacy settings change how data is kept and routed.
- Company policies: OpenRouter privacy policy · OpenRouter terms
Google Gemini (chat models)
- What it receives: For each chat request: the earlier messages in the chat, the site owner's instructions, the text from attached files, any image attached to a message, the model name and any settings. It does not receive your name or your email.
- Why: To run the Gemini chat models that the site owner has turned on.
- On or off: Optional. Used only when the site owner has added a Gemini key. This is a Gemini key, not Google sign-in.
- Free and paid use: Google's rules depend on whether billing is on for the Google Cloud project that holds the key. On free use, Google may use what you send to improve its products. Human reviewers may read it. Google advises against sending personal information to free services. On paid use, Google says it does not use prompts or responses to improve its products. Paid-use logs of prompts and responses are kept for a limited time, to check for policy violations. The default maximum on Google's logs page is 55 days. The site owner's billing setup decides which use applies. The owner can check this in Google's console.
- Where: Google runs servers around the world. On paid use, data may be stored or cached in any country where Google has facilities.
- Company policies: Google privacy policy · Gemini API terms · Google data processing addendum
Google Gemini (image reading and audio fallback)
- What it receives: Each image you upload (PNG, JPEG, WebP, HEIC or HEIF, up to 4 MB) is sent to Gemini, with a fixed instruction to describe the image and copy its visible text. The description is then sent as text to the chat model you pick. If a Gemini model fails, the same image goes to the next model on the list. For audio, Gemini is used only when Groq is not available. The audio is sent with a fixed instruction to transcribe it.
- Why: To turn images into text, so any chat model can use them. To transcribe audio when Groq cannot.
- On or off: Optional. It needs a Gemini key. Without one, image uploads are refused.
- Free and paid use: The same rules as Google Gemini chat above apply. Images can show people and personal details, and Google advises against sending personal information to free services.
- Company policies: Google privacy policy · Gemini API terms · Google data processing addendum
Anthropic (Claude models)
- What it receives: For each Claude reply: the earlier messages in the chat, the site owner's instructions, the text from attached files, image files (JPEG, PNG, GIF and WebP), the model name and the longest reply allowed. HEIC and HEIF images are not sent, but their description is. Anthropic does not receive your name, email or account ID.
- Why: To run Claude for accounts that the site owner has switched on. Claude use is metered and billed to the account.
- On or off: Optional. It needs an Anthropic key saved by the site owner, and Claude turned on for your account. When payments are on, your account also needs a saved card and no failed invoice.
- Use and keeping: Anthropic's commercial terms say it may not train its models on customer content from its services. This site uses Anthropic's API under those commercial terms, not its consumer apps. Anthropic says it keeps standard API inputs and outputs for up to 30 days. It can keep content it flags for safety for up to 2 years. Some Claude models must keep data for 30 days.
- Where: Anthropic stores data on servers in the United States and in other countries outside Europe and the UK. By default, it may run requests in any location.
- Company policies: Anthropic privacy policy · Anthropic commercial terms · Anthropic data processing addendum
Ollama (a model server the site owner can run)
- What it receives: Only if the site owner turns it on. Chat requests then go to the model server the owner sets. By default, that address points to the API server itself, which runs on Railway, not on the owner's computer. So by default no outside company receives the chat text. The requests carry the earlier messages, the owner's instructions, the text from attached files and the model name. No key is sent. If the owner points the site at another company's server, that company receives the chat text, and its own policies apply. Ollama's privacy policy says it collects details about model downloads from its library.
- Why: To let the owner run their own model server and offer its models in the chat.
- On or off: Off unless the site owner turns it on.
- Company policies: Ollama privacy policy · Ollama terms
Cloudflare (website hosting)
- What it receives: Every request for a page or file on the website: the page address asked for, your browser's request headers, and your network address. If request logs are turned on for the website, Cloudflare keeps them for 3 days on the free plan or 7 days on a paid plan.
- Why: To serve the website: the start page, chat, admin, terms and privacy pages. It also redirects plain http to https.
- On or off: Always on.
- Use: Cloudflare's Responsible AI page says Cloudflare may use samples of data that pass through its systems to train its machine learning models. We have not confirmed whether that covers requests to this website.
- Where: Cloudflare says it primarily stores data in the United States and the European Economic Area.
- Company policies: Cloudflare privacy policy · Cloudflare terms · Cloudflare customer DPA
Railway (API hosting)
- What it receives: Everything the browser sends to the API: your sign-in cookie, the values Google returns during sign-in, usernames, passwords (which are hashed before they are stored), chat messages with the text of attached files and any image data, uploaded file bytes while a request runs, and admin actions. It also sees your network address at its edge.
- Why: To run the API server for sign-in, chat, file reading, billing and the admin panel.
- On or off: Always on. The site cannot work without it.
- Logs: The server's log output, including error details, is kept by Railway for a limited time. How long depends on the Railway plan.
- Use: Railway's privacy policy says it may create anonymized or aggregated data from personal data, and use it to improve its services. Its documents do not say whether it uses customer data to train AI models.
- Where: Railway says its main processing is in the United States, on Google Cloud infrastructure.
- Company policies: Railway privacy policy · Railway terms · Railway data processing addendum
Railway Postgres (database)
- What it holds: Everything the site stores: account records (Google account ID, email, name, photo link, username, password hash, Stripe customer ID, role and limits), sign-in sessions (stored as hashes), saved chats, usage counts, Claude billing records, invoices, site settings, and the model provider keys, which are stored encrypted. The database itself is not shared with any other company.
- Why: To hold all site data.
- On or off: Always on.
- Where: The database is hosted by Railway. Railway says its main processing is in the United States, and it also offers other regions. The region is set in the site owner's Railway account.
- Backups and plan: If the owner turns on backups, Railway keeps them for several weeks or months. How long data is kept after a hosting plan ends depends on the plan.
- Company policies: Railway privacy policy · Railway terms · Railway data processing addendum
Cookies
The site sets two cookies:
- session: keeps you signed in for up to 30 days. Guest visits use it too.
- oauth_tx: used only during Google sign-in. It lasts 10 minutes at most.
The site also keeps three things in your browser's storage:
- Your theme choice: light, dark or system. It stays until you clear it.
- The model you last picked in the chat. It stays until you clear it.
- An unsent message, only if you were signed out when you tried to send it. It is kept in that tab until the chat page next loads or the tab closes.
We set no tracking or advertising cookies.
Network addresses
Our server uses your network address in memory only. It uses the address to slow down sign-ups, guest visits and sign-in attempts. These counts are lost when the server restarts. We do not save your address in our database, and we keep no request log that includes it.
Some companies see your address as part of delivering requests. These are Cloudflare, which serves the website, and Railway, which runs the API. Google also sees your address when you sign in, and when your browser loads your profile photo. Model providers see the address of our server, not yours.
Where data is stored and processed
Our database and our API server run on Railway. Railway says its main processing is in the United States. The website is served by Cloudflare, which says it primarily stores data in the United States and the European Economic Area.
Some companies process data in other countries too. Groq stores customer data in the United States. Anthropic and OpenRouter may store data in the United States or in other countries. Google runs servers around the world. Stripe may move data to other countries, including the United States and India. By default, the Ollama address is the API server itself, unless the owner points it elsewhere. Each company's policy, linked above, says more.
Deleting your data
You can delete any saved chat yourself. Click the x button beside it in the sidebar and confirm. Deleting a chat removes its messages and the text of its files. This cannot be undone.
Signing out ends your session and removes its sign-in record.
You can remove Chatbots' access to your Google account in your Google Account settings. That does not delete the copies of your details that we already hold.
The app does not yet let you delete your account, usage counts, Claude billing records or invoice records. For questions about your data, see Contact below.
Changes to this policy
We may change this policy. When we do, we update the date at the top of this page. If we change how we use information from Google, we will tell you on this page.
Contact
Questions about this policy or your data: noicefn01@gmail.com