Before GripIQ gets an answer from an AI model, a question has to go to it first. And that question contains text: your brand context, the content of the context documents you uploaded, sometimes an entire conversation with the assistant. If you work in marketing, you rightly wonder what happens to that text. There are plenty of general promises about AI and privacy. Here, we make it more concrete: which conditions does GripIQ send with every request, and where do they not apply?

What a request contains

A request to an AI model is rarely a single sentence. To give useful answers, a model needs context. At GripIQ, that can include:

  • the Prompts you have monitored;
  • your project's goals and context, with a summary of your context documents, such as a briefing or a brand strategy;
  • your project's personas and the details of your own brand and your competitors, for the weekly Insights;
  • the course of a conversation with the assistant.

That is no side issue. It is exactly the kind of text you don't want to see ending up out in the open. So the question is not whether data goes along, but under which conditions.

The conditions that go with every request

With every request to an AI model, GripIQ sets two conditions for processing:

  • No collection. The request may only be processed by parties that do not collect your text and do not train on it.
  • No retention. For text requests, this goes one step further: the request may only be processed by parties that commit to zero data retention. That means they don't store the request.

These conditions apply to all requests to AI models you deal with in GripIQ: the answers to your Prompts, the weekly Insights and the assistant. They are built into the platform, and don't depend on a setting someone could switch off per project or per feature. If a request can't meet the conditions, GripIQ would rather let it fail than have it processed without them.

AI and privacy: where the limits lie

An honest account of AI and privacy also mentions what does not fall under the strictest condition, and what a condition cannot guarantee.

  • Images for personas. For generating an image for a persona, only the first condition currently applies: no collection and no training. For those requests, no party is available that rules out retention. So the description of a persona may be stored in that case. That is a narrower guarantee than for text requests, and we want to be open about it.
  • A condition remains an agreement. GripIQ requests these conditions and only works with parties that commit to them. What happens inside that party remains an agreement with that party, not proof.
  • What you enter yourself. Everything in a request goes to the AI model that answers it. So don't put anything in it that you don't want processed anywhere.
  • Other parts of the platform. These conditions concern requests to AI models, not other parts of the platform.

That way, you know exactly where you stand, including with the exceptions.

Who can access your data

Separately from the request to the AI model, your data within GripIQ is ring-fenced per project. You can only open the projects you are a member of. Within a project, a team member's role determines what they may change or export; more on this in Roles per team member: sales and management look on without changing anything. So one layer determines who can access the data, the other what happens to a request as soon as a model is involved.

What you can do now

  • Decide deliberately which documents you upload as context documents. What they contain can be passed on, in summarised form, as context in a request.
  • Keep in mind that what you type in a conversation with the assistant goes to an AI model. Don't put anything in it you don't want processed anywhere.
  • Ask other AI suppliers too, outside GripIQ, whether requests are stored or used for training.

Want to see in a demo how we handle your data? Book a demo.

What you can do with the answers themselves is explained in Generative engine optimization: what it is and where to start. You'll find a practical application in How to build a content strategy on the questions AI answers leave open.