Security & privacy · 2026-10-17
Are your emails private when you use AI?
What happens to your emails when AI helps analyse them? Understand consent, storage, model training, providers and privacy before connecting.
You have 200 emails to review. Who gets to see them?
Imagine a long conversation with a company about a missing refund. You want the important dates and commitments, without reading every reply again. AI can help you organise the story. But it is reasonable to ask: does the assistant need to read my emails to do that?
The answer depends on how the product is designed. Privacy is not a single switch. It includes which emails are selected, where copies are stored, who is authorised to access them and whether an external service processes their contents.
Three different things: storing, processing and training
Storing means keeping a copy so you can return to it. Processing means using content for a task, such as answering your question or preparing a timeline. Training means updating a model so that its behaviour may change on future requests.
A provider can process a message without training its model on it. But processing can still involve temporary retention or logs. That is why a meaningful privacy policy explains each step separately.
What MailTrace says about your choices
MailTrace's published privacy policy says Gmail AI processing requires your approval for the case before relevant email evidence is sent to an external AI service. Certain file operations can require additional approval. The policy also explains that saved evidence can remain after an account is disconnected, unless separately deleted.
This is not a promise that the service operator or processing providers are technically unable to access any data. The current product relies on access controls, encrypted connections and other controls described in the privacy policy.
A practical privacy checklist
Before connecting an inbox, check what permission the app requests. Ask whether you can choose the messages and date range, whether external AI processing is optional, which service receives content, and how to remove saved evidence.
For example, you may want an app to organise a timeline but not send its underlying messages to an external AI provider. The difference between those actions should be visible and understandable.
The takeaway
The useful question is not simply 'is it secure?' Ask who can access what, for which purpose, with what permission, and for how long. Read the current MailTrace privacy policy at /privacy for product-specific details.