How to Protect Your Privacy in the Age of AI: Advice from Proton’s CEO
Intro
Using AI assistants and chatbots has become routine for millions of people. The convenience is real: you get quick answers, help drafting emails, summaries of long documents, and creative inspiration. But each query you send to a cloud-based AI service is data that leaves your device. Where it goes, who can see it, and how long it stays stored are questions most users don’t think about — until something goes wrong.
In a recent interview with Spiceworks, Proton’s CEO shared what keeps him up at night when it comes to AI and privacy. It’s not that privacy is impossible, he said. But there is one risk that stands above the rest: the lack of transparency and control over how AI models handle user data.
What happened
The CEO of Proton — the company behind ProtonMail, ProtonVPN, and other privacy-focused services — gave an interview published by Spiceworks on June 4, 2026. He discussed the tension between AI adoption and data protection. His central worry is that many popular AI tools collect user inputs for model training or analysis, often without clear disclosure or meaningful opt-out options. Even when companies promise not to use data for training, the technical architecture may still expose information to third parties or store it in ways that are hard to fully delete.
Proton has long advocated for end-to-end encryption and data sovereignty. In the interview, the CEO argued that the same principles can and should apply to AI services. But currently, most mainstream AI platforms are built on centralized infrastructure where the service provider has full access to the raw text of every user interaction.
Why it matters
For the average person, the risks are not abstract. If you paste a confidential work email into a chatbot to ask for a rewrite, that email text may become part of a training dataset. If you ask an AI for medical advice and include personal symptoms, that information could be stored or later exposed in a breach. A 2025 analysis by a security firm found that several major AI assistants had leaked user conversations due to database misconfigurations. Even if a company has good intentions, human error and malicious actors remain real threats.
What keeps privacy advocates and Proton’s CEO awake is not just the technical possibility of data leakage, but the lack of tools for users to verify what happens to their data. “Trust us” is not a viable privacy policy.
What readers can do
You don’t have to stop using AI. But you can take a few practical steps to limit your exposure:
Prefer AI tools that offer end-to-end encryption. Proton is working on encrypted AI features, and a few other providers now offer similar protections. If the service cannot read your data, neither can anyone else.
Check the terms of service — specifically the data use section. Look for clear statements that your inputs are not used for training. Be wary of vague language like “we may use data to improve our services.” When in doubt, assume your data is being collected.
Limit the sensitivity of what you share. Treat every AI conversation the same way you would treat a public message board. Do not include passwords, banking details, health records, or personally identifiable information unless absolutely necessary and with a trusted platform.
Use separate accounts for different purposes. Keep your personal AI use distinct from work-related queries. Many enterprise AI services offer better data protection guarantees, but only if your IT department has configured them properly.
Delete past conversations. Most AI services allow you to clear your chat history. Make a habit of doing this regularly, especially if the service does not have a clear deletion policy on its backend.
Look for data sovereignty features. Some AI providers now offer the ability to choose where your data is stored geographically. This can help you stay within the legal protections of your home country.
Sources
“Privacy in the AI era is possible, says Proton’s CEO, but one thing keeps him up at night.” Spiceworks. June 4, 2026. (Primary source for CEO’s comments and concerns.)
Proton website and blog for their stance on encryption and data sovereignty.
Various security reports on AI data breaches (general industry knowledge; specific reports vary).