How AmeChat uses your data when you use the AI

When you use the AI in AmeChat, we send only the content needed for a reply to third-party AI providers. This notice states what we send, to whom, and when. Details and transfer safeguards are in our Privacy Policy (Sections 5 and 6). By accepting this notice, you allow the transfers described below.

What we send

• your message to the AI;
• messages from the chat that form the context of the request — in chats with other people, only messages sent after the AI was enabled, including other participants' messages;
• emoji reactions on those messages;
• display names of the people whose messages and reactions are in that context, so the AI can tell who said what;
• images and documents that are part of that context;
• when you generate an image or a video — the prompt and the reference photos, videos, and audio you choose; for a video, also the photo behind each character in it;
• facts from AI memory, if it is enabled;
• your custom AI instructions, if you set them;
• request parameters: the selected model and the current date.

We assemble this context separately for each request.

Who we send it to

To the provider of the model that answers you:

• Anthropic — Claude models
• OpenAI — GPT models and GPT Image generation; also voice-message transcription, embeddings for AI memory, and safety checks
• Google — Gemini models and Gemini image generation
• BytePlus — Seedream image generation and video generation

The current provider list with regions and transfer mechanisms: Sub-processor List.

When we send it

Only when you use the AI:

• you send a message in an AI chat;
• you summon a model or reply to its message in a chat where the AI is enabled — in a direct chat the AI can be enabled only with the consent of both participants; in a group or channel it is enabled by an admin;
• you generate an image or a video — the prompt and any reference photos go to the provider of the selected model;
• you transcribe a voice message — this works in any chat and does not require the AI to be enabled; only the audio of that message is sent.

Apart from these, OpenAI receives content only for safety checks required by law: the prompt, references, and result of every generation, posts and their sound before publication, and profile photos and texts (Privacy Policy, Section 5.1). Your messages to other people are not sent anywhere while the AI is off.

No training

Under our contracts with them, the providers may not use your messages, files, or the AI's replies to train or improve their models, and may not disclose them to anyone. For abuse monitoring they may keep the data of a request for up to 30 days, after which it is deleted. BytePlus keeps a request and its result for up to 180 days, in Malaysia, only when its safety filter flags them.

Your consent

You can decline and use AmeChat without the AI. In a direct chat either participant can turn the AI off at any time. In groups and channels the AI is managed by admins — if you don't want the AI in such a chat, ask an admin to turn it off or leave the chat. You can turn AI memory off in settings at any time. Your acceptance is recorded with the date and version of this notice.