No, your conversations are not used to train the LLMs of our AI providers (OpenAI, Mistral, or others).

Yes, all personal information sent by a customer in their messages is anonymized before it is transmitted to an AI engine.
To learn more, see the section: Anonymizing private data

Sadly, for now AI Agents can manage text only. 

More precisely, here are the limitations in terms of managing media:

• AI Agents cannot receive media (images, videos, audio) or documents (PDFs, etc.) in the customer's query.
• AI Agents cannot analyse media contained in a knowledge base to prepare a response to the customer.
• AI Agents cannot generate an AI response to customers with attachments, media or documents.
• AI Agents cannot generate an AI response to customers in a rich message format such as carrousels, buttons, quick replies, lists, WhatsApp products, etc.

This feature, though a usual one on AI platforms we know and use such as ChatGPT, Gemini or Microsoft Copilot, is currently not possible on messaging channels such as WhatsApp, Messenger, RCS, Instagram DM, Apple Messages for Business, etc. as they don't support text streaming. 

This is a limitation of those channels, independently from Alcmeon.

The text that the AI sends to the customer as a reply can be formatted with line breaks, links and other advanced elements such as bold text, ordered and unordered lists, etc.

Only the AlcmeonChat channel allows advanced formatting of the text in AI responses. Indeed, other channels such as Messenger and Instagram do not allow any text formatting other than line breaks and links in the form of URLs. This is not an Alcmeon restriction.

For AlcmeonChat, here is the accepted formatting, which is based on the simple “Markdown” standard:

codeblock

Note: Tables cannot be displayed in message bubbles, so they are ignored. Similarly, media such as images and videos are handled separately in messages on these channels and never appear in the body of the text.