Your work is not our training data

There is a version of AI development where customers and product are in an extractive relationship: users do work, the system learns from that work, and users receive an incrementally improved product in exchange for surrendering something they didn't know they were giving away.

That model is easy to run and easy to justify. It produces results. It also makes a quiet assumption that we don't accept: that your work belongs, in some functional sense, to us.

It doesn't.

What privacy usually means in this space

The AI industry has developed a particular vocabulary for privacy. Promises tend to cluster around data security — encryption at rest, access controls, not selling data to third parties. These are meaningful guarantees. They are also the floor, not the ceiling.

The harder question is not "who else can see your data?" but "what is your data being used to build?" A system that stores your information securely but feeds it into model training has learned from you without asking. The privacy audit passes. The extraction is still happening.

We want to be direct about our position: askFinz does not train on your work. The documents you research, the questions you ask, the analysis you run — none of that is used to improve the systems that serve other customers. When you close the session, the work is yours.

Why this matters beyond compliance

The practical case for this stance is not just ethical — it is about what kind of relationship makes sense for the work askFinz is built to support.

If you are a professional using askFinz for client research, financial analysis, strategic work, or competitive intelligence, the value of the tool depends entirely on your confidence that the work stays contained. A platform that might absorb your proprietary thinking into a shared capability layer is a platform you would be rational to distrust — or to work around, keeping your most sensitive questions off it entirely.

That self-censorship is the worst outcome. A tool shaped by what you're willing to say to it is a smaller, less useful tool. We'd rather earn the kind of trust that makes people comfortable doing their real work inside the platform.

The harder engineering consequence

Not learning from customer data is not free. There are signal sources we don't have access to. There are patterns we cannot refine through production feedback. Other platforms that do train on user sessions have access to a continuous improvement mechanism we've chosen not to use.

We think that's the right trade. Building better systems without taking people's work as a resource requires more investment in other signal sources, better evaluation methods, and more careful thinking about what it means to improve a product without treating users as contributors to a shared dataset they never opted into.

The work of building responsibly is harder than the alternative. It is also the only version worth doing if the product is going to be trusted with meaningful work.

Where the line sits

Our position is a commitment, not just a policy. Policies can be revised by legal teams in response to pressure. Commitments are part of what you are.

We do store data to make the product function — context within a session, settings and preferences, usage patterns at an aggregate and anonymised level. That is different from learning from the substance of your work. See /trust for the specifics of what we retain and for how long.

The distinction we're drawing is simple: your work is a record of your thinking. It is not raw material for ours.