What Keeps Proton’s CEO Up at Night About AI Privacy — and What You Can Do About It

Intro

Proton’s CEO has long argued that privacy doesn’t have to be sacrificed in the age of artificial intelligence. But in a recent interview with Spiceworks, he admitted there’s one thing that genuinely worries him. It’s not that AI is inherently dangerous, or that encryption is failing. It’s something more specific — and more immediate for anyone using AI-powered tools today.

What Happened

In the June 2026 interview, Proton’s CEO (Andy Yen) discussed the state of privacy in the AI era. While he remains optimistic that strong encryption and privacy-first design can coexist with AI, he pointed to a particular threat that keeps him up at night. According to the coverage, the concern centers on AI agents — software that can act on your behalf, reading your emails, browsing your files, and making decisions with little transparency about how that data is handled.

The exact phrasing from the article is important, and I can’t recreate it verbatim here. But the core warning is this: as AI becomes more agentic (acting autonomously on your behalf), the risk of exposing your most sensitive personal data to companies that don’t prioritize privacy grows sharply. Proton’s CEO reportedly said something along the lines that we’re entering a world where AI will have access to everything — and that’s a problem if the infrastructure isn’t built on a foundation of end-to-end encryption and user control.

Why It Matters

This isn’t a hypothetical concern. If you use a smart assistant, a chatbot that can read your documents, or a tool that summarizes your email inbox, you’re already handing over data that could be used to train models — or leaked in a breach. The problem is that many AI services collect and store that data on their servers, often in ways you can’t fully review or revoke.

The shift to agentic AI means these tools won’t just passively answer questions. They’ll take actions: book appointments, send messages, fill out forms. Each action creates a trail of data that could expose not just what you say, but what you intend. For privacy-conscious consumers, that’s a legitimate reason to stay awake at night — especially if you rely on services like Gmail or Office 365, where AI features are being turned on by default.

What Readers Can Do

The good news is that you don’t have to avoid AI entirely. You just need to approach it more cautiously. Here are practical steps you can take today:

  1. Audit which AI tools have access to your data. Go through your Google, Microsoft, and Apple accounts. Look for permissions that allow AI features to read your email, access your files, or view your photos. Disable anything you don’t actively use.

  2. Use end-to-end encrypted services for sensitive communication. ProtonMail, ProtonDrive, and similar services keep your data encrypted so that even the provider can’t read it. This is especially important if you plan to use AI features within those tools — ensure they are designed with privacy in mind from the start.

  3. Check where AI processing happens. Some tools process data locally on your device (on-device AI), which is much safer than sending everything to a cloud server. For example, Apple’s Siri can now process many requests locally, and some AI keyboards do the same. Prefer tools that advertise “on-device” processing.

  4. Review privacy policies — but focus on data retention and third-party sharing. You don’t need to read every page. Look for two things: how long your data is stored after you delete it, and whether the company shares it with third parties (like AI model trainers). If either answer is vague or unlimited, consider alternatives.

  5. Be careful with AI agents that have “write” access. The most dangerous tools are those that can not only read your data but also act on your behalf. For example, an AI assistant that can send emails from your account or access your calendar. Only grant that level of trust to services you have thoroughly vetted, and turn it off when you don’t need it.

  6. Use separate accounts for AI experiments. If you want to test a new AI service, don’t give it access to your primary email or document store. Create a dedicated account with minimal data, and never use it for anything sensitive.

Sources

  • Spiceworks, “Privacy in the AI era is possible, says Proton’s CEO, but one thing keeps him up at night,” June 4, 2026. (RSS summary suggests the threat relates to agentic AI and data handling.)

This article is based on a single interview, so take the exact phrasing as reported. The actionable steps above are general best practices that apply regardless of the specific warning — they are not direct recommendations from Proton’s CEO unless noted.