Proton, the Swiss company behind Proton Mail, Proton VPN, Proton Pass and Proton Drive, has just had the worst week in its history. Reporting built on leaked Slack messages, meeting audio and interviews with current and former staff alleges that CEO Andy Yen has been forcing an AI-first overhaul on the company, brushing aside security concerns and belittling the employees who raised them. On Friday 30 October, Yen cancelled a round of company-wide AI “re-interviews” that were due to start on 4 November. The Proton AI controversy is nowhere near over, though, because the cancelled interviews were never really the problem.
What happened at Proton
The short version is that Proton told staff they’d be re-interviewed about how they use AI, a leak revealed how hard leadership has been pushing AI internally, and the CEO walked back the interviews a day after the full story landed.
It started on Monday 26 October with an internal email telling employees that selected staff would be re-interviewed between 4 and 11 November to assess how they were adapting to AI. The email described the process as “consequential” and said the outcomes could affect roles, responsibilities and continued employment. Proton responded publicly by insisting it wasn’t carrying out layoffs with AI as cover, saying it was short on talent and still hiring.
Three days later, 404 Media published an investigation into Proton’s internal AI push, drawing on eight current employees, one former employee, Slack chats, meeting recordings and an AI manifesto written by Proton’s CTO, Bart Butler. The detail that set everything alight came from the company’s security Slack channel in late August. An employee asked for a few days to penetration test newly written code before it went live, and Yen reportedly dismissed the request as “bullshit and loser thinking”, telling staff to focus on going faster and to stop hand-crafting so much of their code.
The reporting also describes a manifesto telling engineers their old jobs no longer exist, software installed on staff computers to track how much AI-generated code the company produces, and Yen saying this would be the last year Proton hires engineers at all. Staff described a culture where people with real security concerns felt unable to raise them for fear of being mocked in front of colleagues.
Proton’s response to the AI backlash
At an all-hands meeting on 30 October, Yen admitted to a communication failure on his part and cancelled the re-interviews. In the same breath, he made it clear the direction hasn’t changed, telling staff “I’m still going to push” on AI adoption. Employees later reported that whatever the re-interviews were meant to measure will now be folded into regular performance reviews instead.
Yen also flagged a reorganisation for next month, promising no jobs would be lost to AI, and said Proton’s board is closely monitoring the situation. Butler told staff the CEO had committed to communicating better, and that using AI pragmatically doesn’t mean abandoning security or sound engineering. By his own account, leadership knows it messed up.
Staff reaction was split. Some were glad to see movement. Others described the whole thing as damage control, pointing out that the re-interviews were only scrapped once the story went public, despite concerns being raised internally beforehand. One employee suggested real accountability was unlikely given Yen’s reported majority shareholding. Proton is owned by the Proton Foundation, whose board includes Yen alongside Sir Tim Berners-Lee, and how that board responds over the coming weeks will tell us a lot.
Users, meanwhile, have been giving out in huge numbers. Yen’s own public defence on Proton’s community forum was heavily downvoted, there have been open calls for him to resign, and plenty of paying customers say they’ve cancelled or won’t renew.
Why this hits Proton harder than any other tech company
Proton doesn’t compete on price, design or convenience. It competes on being more trustworthy than the companies its customers are trying to escape, and that’s precisely what this story damages.
Here in Ireland, where most of the big US platforms have their European headquarters and the Data Protection Commission routinely fines them, the appeal of a Swiss privacy alternative is obvious. People don’t pay for Proton Mail because it’s slicker than Gmail. They pay because they don’t want an advertising company reading their inbox. When a Google or a Microsoft forces AI into its workflows, nobody’s surprised. When the company that built its entire brand on being the careful, principled alternative does the same thing and labels a security request “loser thinking”, it lands very differently.
To be clear, using AI isn’t the sin here. Proton already makes its own AI assistant, Lumo, and plenty of engineering teams use AI tools responsibly. The problem is the manner of it. A company holding your emails, your passwords, your files and your VPN traffic is reportedly treating caution as an obstacle to speed. That’s the one trait its customers are paying it not to have.
Is Proton Mail still safe to use?
Yes, for now. Nothing in the reporting suggests Proton’s encryption has been broken, that there’s been a breach, or that anyone’s data has been exposed.
The concern is about the company ethos and future rather than the present. End-to-end encryption means Proton can’t read your messages, but the apps doing that encrypting are still software, and software is only as reliable as the process behind it. If code is being shipped faster with less human review and less testing, the risk of a mistake slipping through goes up. Should you be worried about that? A little, yes. Proton’s apps are open source and have been independently audited in the past, which helps, but the reassurance users need now is evidence that rigour hasn’t been quietly traded for velocity.
What Proton needs to do to win back trust
Another statement won’t cut it. Proton needs to show its work.
That means publishing a clear policy on where AI-generated code is and isn’t allowed, particularly anywhere near encryption, authentication or account security. It means committing publicly to human review and penetration testing for security-critical code, and backing that up with fresh independent audits whose results are published in full. It also means decoupling AI usage from performance reviews entirely, because moving the same measurement into a quieter process isn’t a change of direction. Finally, it means leadership owning the culture problem staff described, not just the communication problem Yen has admitted to.
None of that is unreasonable for a company that asks people to trust it with the most sensitive parts of their digital lives.
I don’t have a dog in this fight, but it’s worth acknowledging that some have called for Yen to step down too.
Should you leave Proton?
Not in a panic. If you’re on a bundle covering Mail, Pass, VPN and Drive, switching everything is a serious job, and nothing has gone wrong with your data. Watching what Proton does over the next few months is a perfectly sensible position.
That said, if your renewal is coming up, shopping around is fair. Tuta and Mailbox.org are both established, privacy-focused email providers based in Germany, Bitwarden is a strong open-source password manager (I host my own) , and Mullvad has a solid reputation in the VPN space. Spreading your services across a few providers also means one company’s bad week doesn’t affect everything at once. Whatever you decide, make sure you know how to export your data from Proton so you’re never locked in.
Proton’s real dilemma is that it can’t win an AI race by becoming indistinguishable from the companies it was set up to challenge. It can absolutely use AI. What it can’t do is ask privacy-conscious people for their trust while treating caution as a weakness.
