How to do it:

What’s the problem?

While there are many issues with AI, from content rights to environmental damage, in the context of absorbing your email there’s a particularly glaring problem: privacy leakage. AI doesn’t have a concept of personal property. AI “leaks” ideas across context—that’s how it learns. Could it helpfully enhance a note to your mom with that colorful but unflattering nickname from a rant to your best friend? You’d catch that, right? That’s about the most benign example we can come up with—if you have genuine security issues the risks are far higher than an embarrassed apology.

Google says it isn’t using your email to train Gemini AI, which may mean the AI tool can’t be used to inadvertently spy on your correspondence, but it would take technical privacy protocols to prevent leakage within the Suite. There are no indications Google has taken such steps, and some unrelated changes to their AI policies don’t reassure us. If you share our concerns, you can turn off the feature, but you will lose some tools, including decades-old functions like Spellcheck:

• Go to All Settings in Gmail
• Under General, scroll to Smart features and personalization
• Uncheck the box
• Then go to Google Workspace → Smart features → Manage Workspace and toggle off both smart-feature settings

If this is the final straw for you, we recommend looking into Proton Mail as an alternative. Like Gmail it offers a free basic account for personal use. Unlike Gmail, it is a privacy-first tool governed by European privacy law. Wired has a thorough, recent review about moving from Gmail to Proton.

Proton does offer an AI fueled writing assistant, if this is a function you’ve become reliant on, but you can choose to run it locally on your own computer, avoiding privacy concerns. The Independent has more information about Proton’s advanced functionality including AI tools. If you’re looking for a privacy-forward option to replace Gmail, Proton is the recommendation of nearly every security-minded geek we know, and moving to a European company provides a solid level of protection against U.S. government surveillance.

Note: Paid, domain-branded gmail accounts have also integrated AI, so if you use gmail to run your business domain email this applies to you, too. We simply don't know what it will mean. Could the AI take a confidential pitch for one client and "leak" it into the inquiry a new client sends? Will it have the legal right to writing snippets your editor revised? There are few (if any) legal protections to what AI can do with material you allow it to train on.