What Proton’s CEO Says About AI Privacy – and the One Risk You Need to Watch

Most people who use AI assistants like ChatGPT, Gemini, or Copilot never stop to ask: where does my conversation go? The question gets more urgent as companies continue to train models on user data, sometimes without clear consent. In a recent interview with Spiceworks, Proton’s CEO Andy Yen made a case that privacy in the AI era is possible, but he also pointed to a specific threat that keeps him up at night.

That threat, according to the article, is not a new form of surveillance or a fancy zero‑day exploit. It is something far more mundane: the risk that your inputs to AI tools become part of the training set for future models, and that sensitive personal or business information leaks out as a result. Yen emphasized that many users have no idea how their data is being processed after they hit “send”.

This article breaks down what that risk really means and what you can do about it today.

What happened

Andy Yen, the founder and CEO of Proton (the company behind Proton Mail, Proton VPN, and Proton Drive), gave an interview to Spiceworks in June 2026. In the conversation, he acknowledged that privacy online is under pressure from the rapid adoption of generative AI. However, he said privacy is still achievable – but only if users take deliberate steps. The one specific concern he highlighted is the way AI service providers may log, retain, and use conversations for model training and improvement.

Proton itself has started offering its own AI assistant that runs on encrypted infrastructure, but Yen’s warning applies to the broader ecosystem, especially free‑tier services that rely on data collection for revenue.

Why it matters

If you paste a draft contract into an AI chatbot to get a summary, or if you ask a chatbot to rewrite a sensitive email, you might be handing that information to a company that could later reproduce it in a response to another user. Several incidents have already been reported where AI models regurgitated private details from their training data. For professionals handling legal, medical, or financial information, this is not hypothetical. Even for ordinary consumers, sharing personal stories, passwords (by mistake), or location data can lead to real harm.

The risk is compounded by the fact that many people treat AI tools like search engines – they ask anything without thinking about retention policies. In most cases, the terms of service allow the provider to use the data to improve the model, unless you manually opt out. The opt‑out process is often buried in settings and varies by platform.

What readers can do

You can protect yourself without giving up the convenience of AI. These three steps are practical and can be done in a few minutes:

  1. Check and change data sharing settings for each AI service you use.

    • For ChatGPT (OpenAI): go to Settings → Data Controls → disable “Improve the model for everyone” (this stops your conversations from being used for training).
    • For Google Gemini: visit your Gemini activity settings and turn off “Gemini Apps activity”.
    • For Copilot (Microsoft): sign in, go to Privacy → “Responsible AI” and turn off data collection.

  2. Use AI tools that run locally or encrypt your data.
    Many consumer‑friendly models can now run on your own device (e.g., Llama, Mistral, or Phi via apps like LM Studio). If that’s too technical, look for services that promise end‑to‑end encryption and do not use your data for training. Proton’s own AI assistant is one example, but there are others like Brave’s Leo (privacy mode) or DuckDuckGo’s AI chat, which does not store conversations.

  3. Treat AI conversations like public messages.
    Until you know exactly how a service handles data, assume anything you type could be seen by a human reviewer or included in a model update. Do not paste passwords, bank details, medical records, or trade secrets. Use a separate, disposable account for each major service you test.

Sources

  • “Privacy in the AI era is possible, says Proton’s CEO, but one thing keeps him up at night” – Spiceworks, June 2026. Link to article
  • OpenAI ChatGPT settings guide (support.openai.com)
  • Google Gemini privacy settings (myactivity.google.com)
  • Microsoft Copilot privacy information (support.microsoft.com)
  • DuckDuckGo AI Chat privacy policy (duckduckgo.com)

If you use AI tools, you don’t have to choose between convenience and privacy. But the choice won’t be made for you – you have to act on it.