Is my data used to train AI? What does the AI see?
Last updated
Plans: Free: not included · Pro: included · Scale: included
No. Replavia’s AI providers do not train their models on what we send, and the AI sees only what a reply needs, with emails and phone numbers removed.
Open Web knowledge →Why this matters
Your prices, your customers’ messages and your followers’ comments are not material for someone else’s AI. Before you switch AI on, you should know exactly what leaves Replavia and what happens to it.
Whose AI account it runs on
The AI runs on Replavia’s own accounts with the AI providers named in our Privacy Policy: OpenAI and Anthropic.
You never add your own AI key, and you cannot. This is deliberate. One agreement between Replavia and each provider then covers every customer. It also means there is no AI bill on top of your plan.
Is it used to train AI?
No. The providers act as our processors under their API terms. They process what we send on our instructions, not for their own purposes. They do not use it to train their models.
Your knowledge library is not training either. The AI is handed the matching facts each time it writes a reply. Nothing is built into a model.
What the AI sees
Only what each feature needs:
AI comment replies
- What is sent
- The comment being answered, plus your business details and facts.
AI DM assistant
- What is sent
- The new message and a few recent messages from that one chat.
Knowledge library
- What is sent
- The text of a page or document you add, once, to pull out facts.
| Feature | What is sent |
|---|---|
| AI comment replies | The comment being answered, plus your business details and facts. |
| AI DM assistant | The new message and a few recent messages from that one chat. |
| Knowledge library | The text of a page or document you add, once, to pull out facts. |
Before a comment or a DM is sent, email addresses and phone numbers in it are found and replaced.
What the AI never sees
- your contact and lead records, or your lead exports
- the email address or phone number saved on a lead
- your notes on a lead in a connected CRM
- messages from chats the assistant is not handling
- your payment details
- your Instagram login token
How long it is kept
- At the AI provider. A provider may keep a request for a short time for abuse monitoring. For OpenAI this is up to 30 days. After that it is deleted, unless the law requires otherwise.
- At Replavia. We do not keep the prompts we send.
- Comment replies. The comment text and the commenter’s username are stored encrypted. Once a reply is finished and 30 days have passed, Replavia deletes the comment, the commenter’s details and the reply text.
- Knowledge. Replavia keeps only the facts it saved. The page or file itself is not stored.
- Usage counts. Counts and estimated cost are kept for up to 24 months, to apply your allowance.
Your facts stay with the right brand
Knowledge sources belong to one Instagram account and are only used for that account. Settings you leave blank for an account use your shared settings, so give each brand its own offer and FAQ.
Tips for safe facts
- Put only public business information in your knowledge library: prices, hours, policies, delivery.
- Do not upload customer lists, private notes or passwords. The AI does not need them.
- Remove a source when it is out of date. Future replies stop using it straight away.
Good to know
- AI replies and the AI DM assistant are on Pro and Scale only.
- The providers may process data outside your country. The Privacy Policy explains the protections used.
- See AI safety checks for what stops a bad reply from posting.