Proton, the Swiss company behind the encrypted email service Proton Mail, is calling selected employees back for interviews about how they use AI, and a leaked email says the outcome can affect whether they keep their jobs.
The company sells privacy to people who don't trust Big Tech's AI. Inside, it is grading staff on how far they've adopted it.
Most companies find out who isn't using AI by checking a license dashboard. Proton is sitting people down and asking them.
The company told employees on Monday that it will "re-interview" staff about how they use AI in their work, according to an internal email obtained by 404 Media. The sessions run November 4 to 11. The email is blunt about the stakes. "These re-interviews are consequential," it reads.
Proton had roughly 500 employees as of 2024. The email says not all of them will be called in, and that any decision affecting someone's employment will be communicated separately. Proton didn't immediately respond to 404 Media's request for comment.
Interviewed by someone who isn't your boss
The design is the telling part. Employees won't be assessed by their own managers. Interviewers come from outside each person's team and reporting line, drawn from a pool of leaders the email describes as strong AI users who have agreed on how to score people.
That takes the person who knows an employee's work best out of the room. A manager might know that a support lead's best quarter had nothing to do with a chatbot. The interviewer from the AI pool doesn't know that, and isn't being asked to find out.
The email says Proton needs people who can turn AI into better products and faster execution, and who meet the change with "curiosity, judgment, and urgency." It doesn't say, at least in what has been reported so far, what a failing answer sounds like or how many jobs are in play.
Proton has been heading here for two years
The direction shouldn't surprise anyone at Proton. The speed might. CEO Andy Yen has been steering the company toward AI since at least July 2024, when Proton put a writing assistant called Scribe into Proton Mail. His reasoning then was practical: customers were going to use AI tools anyway, so they might as well use one that didn't leak their data, he told TechRadar. A year later came Lumo, a chatbot that keeps chat logs encrypted and that Proton says never trains on user data.
Those were product bets, aimed at customers who wanted AI without the surveillance. The re-interviews turn the same bet inward. Proton has decided the people building those products need to be heavy users themselves, and it's prepared to test them on it.
The mandate playbook keeps getting stricter
Proton joins a run of chief executives who have turned AI use from a suggestion into a condition of the job. Each version has been a little blunter than the one before.
Shopify's Tobi Lütke went first in April 2025 with a memo calling AI use a baseline expectation. He added AI questions to performance reviews and made teams prove AI couldn't do the work before asking for headcount. "I don't think it's feasible to opt out of learning the skill of applying AI in your craft," he wrote, per eMarketer. Duolingo's Luis von Ahn followed weeks later, declaring the company AI-first and saying it would gradually stop using contractors for work AI could handle.
Coinbase dropped the subtlety that August. Brian Armstrong said on a podcast that he gave engineers a week to get set up on Copilot and Cursor, then met with the ones who hadn't. Those without a good reason were fired, TechCrunch reported. He admitted the approach was heavy-handed.
Proton's version goes further than any of them in one respect. Shopify and Duolingo folded AI into reviews employees already sit through. Coinbase checked whether engineers had installed a tool. Proton is running a separate, company-wide evaluation with a deadline, judged by outsiders, with employment named as a possible outcome. Installing the software won't get anyone through it. They'll have to talk convincingly about how they use it.
The timing isn't subtle either. Proton's email landed the same day Microsoft offered buyouts to about 60,000 employees as AI spending reshapes its budget, and less than a week after Barclays put developers on a cut list three weeks into its Claude Code rollout.
A hard sell for a privacy brand
Most of Proton's customers came for one reason. They didn't want their email read, their files scanned or their conversations used to train someone else's model. A good share of its audience is skeptical of the AI industry on principle.
Proton has tried to meet those users halfway. Its own 2024 survey found more than three quarters of its business users were interested in generative AI, and most of them worried about where their data would end up. Scribe and Lumo were built for exactly that tension.
Grading employees on AI fluency is harder to square with the brand. Plenty of Proton's staff are the same kind of privacy-minded technologists as its users. Some presumably joined because Proton was the company that didn't chase every Silicon Valley trend. Being told their jobs may hinge on proving they've embraced one is a different deal from the one they signed. Customers paying attention will notice that too.
What to watch after November 11
First, whether Proton says anything publicly. A company whose brand rests on being straight with people has so far let a leaked email do the talking.
Then the numbers. How many people get called in, and how many come out with a different role or none at all. If the count is small, this reads as an aggressive internal push. If it's large, it's a layoff run through an interview process, and other companies will be watching to see whether it works.
Proton built its business on the idea that people should be able to say no to Big Tech. Starting November 4, some of its own employees will find out whether they can say no to AI.
