Yes, You Can Use AI Without Sacrificing Privacy: Here’s How

In a recent interview with Spiceworks, Proton’s CEO made a claim that might surprise anyone who has read the fine print of popular AI chatbots: privacy in the AI era is possible. But he also admitted there’s one thing that keeps him up at night—and it’s not technical limitations. It’s the quiet resignation of users who assume they have no choice but to trade their data for convenience.

This isn’t a new problem. What’s changed is how easy it is to hand over personal information without realizing it. Every prompt you type into a free AI tool, every document you upload for summarization, every image you generate—it can all become training data, or worse, get shared with third parties without your knowledge.

The good news is that you don’t have to swear off AI to protect yourself. You just have to be deliberate about the tools you use and how you use them.

What keeps Proton’s CEO up at night?

According to the Spiceworks article, the CEO’s primary worry isn’t encryption or policy. It’s user complacency. Many people, he argues, have accepted that giving up privacy is the price of using something like ChatGPT or Gemini. They no longer bother looking for alternatives or adjusting settings, because they assume no better option exists.

That assumption is becoming less true every quarter. But unless users demand change and make informed choices, even well‑intentioned privacy tools will struggle to gain traction.

Why it matters: what’s really at risk

When you use an AI assistant, you’re typically sending your inputs to a company’s servers. That company may:

  • Store and analyze your conversations to improve its models.
  • Share aggregated (or sometimes not‑so‑aggregated) data with advertisers or partners.
  • Retain your prompts even after you delete your account, depending on their policy.

This matters beyond abstract “privacy.” If you use AI for work, you might expose confidential business ideas. If you use it for health questions, you could inadvertently hand over sensitive medical information. Even casual usage—asking for recipe suggestions or travel itineraries—creates a record that can be linked to your identity.

What you can do: practical steps, not paranoia

You don’t need to become a hermit to use AI responsibly. Here are concrete actions you can take today.

1. Vet tools for privacy first
Look for three signals in any AI service you consider:

  • End‑to‑end encryption for your conversations (not just in transit, but at rest).
  • A clear no‑log or minimal‑log policy—the service should state explicitly that your data is not used for training.
  • Options for local processing where the AI runs on your device rather than in the cloud.

Examples that meet these criteria (to varying degrees) include DuckDuckGo’s AI Chat, Brave’s Leo, and Perplexity’s Pro mode with “limited data collection.” Proton itself, having launched a privacy‑focused AI assistant for its users, encrypts queries so that even Proton can’t read them. Always verify the current policy, as offerings change.

2. Configure existing services to limit data collection
If you still use a major provider like ChatGPT or Google Gemini, dig into your settings:

  • Turn off chat history and model improvement toggles.
  • Avoid uploading files unless absolutely necessary.
  • Use dedicated temporary sessions (e.g., incognito mode in the browser) to reduce cross‑session tracking.

None of these steps make you fully private, but they significantly reduce the surface area.

3. Be strict about what you share
Treat every AI prompt like a message you might later see on a billboard. Don’t include:

  • Your full name, address, or phone number.
  • Passwords, API keys, or financial account details.
  • Work‑specific project names or client data.

If you must use AI for sensitive work, consider writing a sanitised version of the input. Remove proper nouns and vague specifics before hitting send.

4. (Bonus) Try self‑hosted or open‑source models
For the technically inclined, running a model like Llama 3 or Mistral on your own machine gives you total control. No data ever leaves your computer. Tools like Ollama or LM Studio make this surprisingly easy on a modern laptop—no cloud required.

Sources

  • Spiceworks interview with Proton’s CEO (June 2026) – summary via Google News (no full text available at time of writing).
  • DuckDuckGo AI Chat privacy policy (duckduckgo.com).
  • Brave Leo privacy FAQ (brave.com).
  • Proton’s official documentation on AI assistant privacy (proton.me).

Note: Product policies change; always check the latest version before relying on a service for sensitive data.