How we use AI

In short

The Instagram assistants built with InstaForge are AI systems. They always tell people that at the start of a conversation, a person can always take over, and nobody's data is used to train AI models.

Who runs InstaForge

Contact
https://instaforge.app

What the AI does

• writes replies to messages and comments, using only the information the business gave it (it can be set to say "I don't know" rather than guess);
• keeps a short summary of facts someone shared, so it does not ask again (where the business asks for permission first, only after the person agrees);
• turns voice messages into text and describes photos people send, so it can answer them;
• helps the business's team: drafts post captions, suggests answers to questions it could not answer, and turns a website into a knowledge base.

Bookings, deposits, reminders and fixed answers (for example prices) follow rules the business set, not AI judgement.

Being open about it

Every assistant tells people, before its first reply, that it is an automated assistant using AI and which business it acts for. It says so again when the conversation comes back to it from a person, and after a long silence. It never claims to be a person. This follows the EU AI Act (article 50), Meta's messaging rules and the AI providers' terms.

People stay in charge

• Anyone can ask for a person, and the business's team gets an alert.
• The team can read every conversation, step in and reply themselves at any time.
• Questions the assistant could not answer are collected so the team can answer them properly.
• Every change to an assistant is saved and can be undone.
• AI-drafted captions are only published after someone from the team reviews and schedules them.

Limits

AI can be wrong, can misunderstand, and can sound more certain than it should. The business is responsible for what its assistant says and should check its conversations. The assistants are not meant for medical, legal or financial advice, and they make no decisions with legal or similarly significant effects on anyone.

Which AI models

Each business chooses the AI provider for its assistant, for example OpenAI, Anthropic, Google or DeepSeek. The assistant's own privacy page names the one it uses. The provider receives the messages the assistant has to answer, and nothing else. OpenAI, Anthropic and Google's paid services do not use this data to train their models. DeepSeek processes data in China and may use it to improve its models, so we advise against it for people in the EU.

Risk level

Under the EU AI Act, these assistants are not a high-risk use of AI: they answer customer questions, offer appointments and collect requests. InstaForge provides the system and the business that uses it is responsible for how it uses it. Both follow the transparency rules that apply to AI that talks to people.

Questions or problems

If an assistant said something wrong or inappropriate, tell the business, or write to https://instaforge.app. Our privacy policy explains how personal data is handled.