The Cardinal Brief

We will never train AI on your email. Here is what that means in practice.

Somewhere in the terms of service of a surprising number of tools, there is a sentence that means: your content makes our product smarter for everyone. For a lot of software, that trade is defensible. For a tool you hand your email to, I think it is disqualifying.

So Inbox Cardinal starts from a hard commitment: we will never train AI models on your email or calendar content. Not to improve our ranking for other users, not anonymized, not aggregated, not opt-out. Never.

Here is what that means mechanically, in plain language.

When Inbox Cardinal scores a message, an AI model reads it once to answer narrow questions: does this look urgent, is someone waiting on a reply, is this bulk mail. That is called inference, and it is fundamentally different from training. Inference uses a model; training changes one. Your message influences your score and nothing else. It does not flow into any pool of learning material, and no future version of any model is smarter because your mail passed through.

The commitment only means something with the architecture behind it, so here is the rest, briefly. Everything is encrypted in transit and at rest. Consent is per account: you approve each mailbox individually, we request the minimum access each feature needs, and we never ask for organization-wide access to anyone's company. Your content flows directly between your providers and us, with no third-party middleman in the path. And disconnecting an account purges its data. Disconnect means delete, not archive.

One honest distinction, because trust is built on precision: Inbox Cardinal does learn your priorities over time, and that personal tuning is on our roadmap. But learning which senders you consider important is preference data about you, kept for you, and it is worlds apart from using the content of your mail to train models that serve everyone else. The first makes your assistant yours. The second quietly makes you the product.

If you take one thing from this post, make it a habit for evaluating any tool that asks for your inbox: find the sentence about training. If you cannot find a clear answer in plain language, that is your answer.

Ours is on the label, and it will not change: never. The rest of our commitments live on the trust page, in plain language, and hard questions are welcome at hello@inboxcardinal.com.

Join the beta