How to Protect Your Privacy When Using AI: Tips from Proton’s CEO

Generative AI tools are now part of daily life for millions of people. Services like ChatGPT, Microsoft Copilot, and Google Gemini are used for writing, research, coding, and even personal advice. But each prompt you type in can become data that the company uses for training, analysis, or advertising. Privacy concerns have grown alongside adoption.

Proton AG, known for its encrypted email and VPN services, has been vocal about this tension. In a recent interview with Spiceworks, Proton’s CEO laid out where the real risk lies and what users can do about it.

What’s the core problem? Centralized data hoarding

The CEO’s main worry is not that AI itself is dangerous. It’s that so much personal data flows into the hands of a small number of companies. As he put it, roughly: “The biggest risk is that we’re handing all our data to a few companies who can use it without our knowledge.”

Most AI tools operate on a centralized model: your input is sent to a remote server, processed, and stored. That data can be used to fine‑tune models, profile users, or—in some cases—sold or shared. Even when a company promises not to read individual prompts, metadata and usage patterns still accumulate.

For an everyday user, the problem is threefold:

  • You lose control over your own information once it enters the system.
  • The scope of collection is often unclear—company privacy policies differ, and some silently change their terms.
  • Regulatory gaps remain: many jurisdictions have not yet enforced strong AI‑specific data protection laws.

Practical steps you can take

The good news is that you do not have to stop using AI altogether. A few deliberate choices can greatly reduce your exposure.

1. Use privacy‑first AI services when possible.
Proton itself offers an end‑to‑end encrypted email service (Proton Mail), a VPN (Proton VPN), and cloud storage (Proton Drive). Last year it launched Proton Scribe, an AI writing assistant that runs on device for some tasks. Look for tools that process data locally or at least encrypt it before sending. Open‑source models like Llama or Mistral can also be run on your own hardware, though this requires some technical comfort.

2. Read—or at least glance at—the terms of use.
Pay special attention to sections about “training data,” “aggregate usage,” and “data retention.” Some services allow you to opt out of having your conversations used for training. Do that check box now—it only takes a minute.

3. Disable data sharing in settings.
ChatGPT, for example, lets you turn off “improve the model for everyone” in your account settings. Copilot in Microsoft 365 has similar toggles for telemetry. Make sure these are off.

4. Never share sensitive personal information in a prompt.
Treat every AI chat like a conversation in a public square. Do not enter passwords, health records, financial details, or anything you would not want posted online. If a tool asks for that info, stop and find another way.

5. Limit your use of AI for personal queries.
It can be tempting to ask an AI for relationship advice, legal analysis, or a diagnosis. But those inputs become part of a permanent record. Use specialised, privacy‑respecting services for such needs instead.

How the industry is responding

Some steps are being taken at the industry and regulatory levels. Proton’s model of end‑to‑end encryption is being adopted by a small but growing number of competitors. Open‑source AI models allow individuals to run their own instances, and companies like Apple have started processing certain AI functions on‑device.

Regulation is slower. The EU’s AI Act has introduced risk categories, but enforcement is still ramping up. Other regions have yet to pass strong laws.

The CEO emphasized that user pressure matters. When people demand clearer privacy policies and opt‑out options, companies gradually respond.

You don’t have to choose between AI and privacy

Using AI does not mean you have to surrender your data. With awareness and a few consistent habits, you can gain the benefits of automation and assistance while keeping your personal information under your control.

The real shift needed is cultural: companies should treat user data as the user’s property, not a resource to be mined. Until that changes, individuals must stay cautious. The tools are powerful—but so is your ability to decide what you share.

Sources:

  • Spiceworks interview with Proton CEO, published June 4, 2026.
  • Proton AG official service descriptions (proton.me).
  • ChatGPT privacy settings and Microsoft Copilot privacy documentation (accessed June 2026).