Why We Enabled Zero Data Retention for All Our AI Calls

When you type your startup idea into PreVibe, you hand us something valuable. It may be the seed of your next company. It is unreleased intellectual property. We think it deserves real protection, not just a promise in a policy page.

That is why every AI call PreVibe makes now runs with zero data retention. Here is what that means, why it matters, and exactly what we changed.

What zero data retention means

Every AI product sends your text to a model provider. That is how the technology works. The question is what happens to your text after the model answers.

By default, many providers keep prompts for a while. They store them in logs. Some hold them for days or weeks for abuse review. Some use them to train future models unless you opt out.

Zero data retention (often shortened to ZDR) flips that default. The provider processes your request, returns the answer, and deletes the data. Your prompt is not stored after the request completes. It is not used to train a model. There is no copy sitting in a log file months later.

Why this matters for founders

Most AI privacy debates focus on personal data. For founders, the bigger risk is your idea itself.

Your idea is your head start

An unvalidated idea has one asset: nobody else is working on it yet, or at least not many. Every copy of your idea that sits on someone else's server weakens that head start. Retained prompts can end up in training data, in analytics, or in front of staff during a review. None of that is likely, but the cost if it happens is high. The cheapest fix is to make sure the copy never exists.

Stored data can leak

Data breaches happen to careful companies too. A provider cannot leak what it never stored. Zero data retention shrinks the attack surface from "every prompt ever sent" to "requests in flight right now".

Fewer copies, simpler compliance

Under GDPR and similar laws, every place your data lives is a place you must account for. When AI providers keep nothing, the data map gets shorter and the guarantees get easier to keep. That protects you, and it keeps our own privacy promises honest.

What we changed

We did not just tick a checkbox. We enforce zero data retention at three layers:

1. Every request asks for ZDR. Each AI call PreVibe makes carries an explicit zero-data-retention flag. This covers report generation, the verdict, the research chat, and web search grounding. All of it.
2. We pin the providers and block fallbacks. AI gateways route requests to many possible backends, and the list can change under your feet. We lock each call to a named list of providers verified for zero retention, with fallbacks disabled. If the ZDR route is ever unavailable, the request fails instead of quietly going somewhere else.
3. Account-level ZDR is the backstop. On top of the per-request flag, we enabled zero-data-retention settings at the account level with our AI gateway. Even if a code change ever dropped the flag, the account setting would still hold the line.

We also rewrote our privacy policy to name the actual services involved: OpenRouter routes our AI requests, Microsoft Azure runs model inference, and Perplexity powers web search. No vague "trusted third parties" wording. You can see exactly who touches your data and in what role.

What this does not change

Zero data retention is about our AI providers, not about your account. Your reports are still saved in your dashboard, because that is the product: you come back, reread the research, and ask follow-up questions about it. That data lives in our database under your account, and you can delete any research whenever you want.

In short: we keep what you asked us to keep, and the AI providers keep nothing.

The takeaway

PreVibe exists so you can validate an idea before you spend months building it. That only works if validating the idea does not expose it. Now the whole pipeline honors that: your idea goes in, your verdict comes out, and no AI provider keeps a copy in between.

Got questions about how we handle your data? Check the privacy policy or reach out. We are happy to go into detail.