Proton CEO Says AI Privacy Is Possible—But Here’s What Still Worries Him

Proton’s chief executive, Andy Yen, recently told Spiceworks that he believes privacy in the artificial intelligence era is achievable. The company, known for its encrypted email, VPN, and cloud storage services, has been positioning itself as a privacy-first alternative to big tech platforms. Yet in the same interview, Yen admitted that one specific issue still keeps him up at night: the way AI tools are quietly being integrated into nearly every digital service, often without giving users meaningful control over their data.

What Yen described is not a distant hypothetical. It is already happening. Many popular AI assistants, writing tools, and image generators run on servers that collect and store user inputs to refine their models. Even when a service claims not to use personal data for training, the underlying infrastructure often exposes more information than most people realize—metadata, usage patterns, or the content of queries themselves.

The core concern: data extraction at scale

The worry, as Yen explained, is not that AI itself is dangerous, but that the current model of AI deployment makes widespread data extraction almost invisible. When you use a free chatbot, paste a document into an online summarizer, or ask a voice assistant to read your calendar, you are handing over data that could be logged, analyzed, and potentially reused. The systems that power these tools are often run by companies whose business model depends on monetizing user information. Proton’s CEO framed this as a structural problem: the incentives are misaligned, and regulation has not caught up.

That concern is especially sharp because many users do not realize how much they are sharing. A single AI query might contain sensitive personal details, confidential work notes, or private conversations. Once that data leaves your device, you lose practical control over how it is stored, who accesses it, and whether it is later used to train a competing product or sold to a third party.

Why this matters for everyday users

For the average person, the practical consequences can range from annoying to damaging. AI tools are increasingly used for tasks like drafting emails, summarising medical reports, or generating financial advice. If those inputs are not properly protected, a leak or misuse could expose private health information, business strategies, or personal correspondence. Even if no breach occurs, the long-term accumulation of user data into large models can erode privacy in subtle ways—enabling profiling, targeted manipulation, or discrimination.

Proton’s stance is that end-to-end encryption and local processing are the only reliable safeguards. But most mainstream AI services do not offer either. Users are left to choose between convenience and privacy, and many do not know how to evaluate the options.

Practical steps you can take now

You do not need to abandon AI tools entirely to protect your privacy. A few concrete changes can make a meaningful difference:

  • Use AI services that process data locally when possible. Some writing assistants and image editors run on your own device, sending no data to external servers. Open-source models are increasingly capable and can be run on a laptop or phone.
  • Choose privacy-focused providers. Services like Proton’s own AI features (where available) are designed to process data without storing or sharing it. Other encrypted alternatives exist for specific tasks.
  • Review permissions and terms. Before signing up for a free AI tool, check what data it collects and whether it reserves the right to use your inputs for training. If the terms are vague, assume the worst.
  • Separate sensitive tasks from general AI use. Do not paste confidential information—passwords, financial records, private correspondence—into a public-facing chatbot. Treat AI queries the way you would treat a conversation in a public space.
  • Use a privacy-oriented browser or extension that blocks tracking and limits data sharing. Combined with a VPN like Proton VPN, this can reduce the metadata that AI services can collect about you.

None of these steps are foolproof, but together they reduce your exposure while still allowing you to benefit from AI. The key is to treat each interaction as a data sharing decision, not a casual query.

Sources

  • Privacy in the AI era is possible, says Proton’s CEO, but one thing keeps him up at night – Spiceworks (June 2026).
  • Proton official blog and documentation on AI and privacy (various).
  • General guidance from EFF and other digital rights organizations on AI data handling practices.