Proton, the Swiss company behind Proton Mail, was about to put almost its whole staff through an AI exam.
According to 404 Media and the Times of India, employees were told they would be "re-interviewed" between November 4 and 11 about how they use AI. The company email said: "These re-interviews are consequential." Outcomes could change roles, responsibilities and continued employment.
Then, on October 30, CEO Andy Yen cancelled them.
What happened inside Proton
404 Media's investigation, based on eight current and one former employee, Slack chats and meeting audio, described a hard AI push. When an employee in the security channel asked for a few days to pentest new code before it went live, Yen replied: "That's bullshit and loser thinking." He told the team to "stop lovingly handcrafting all our code and use AI more."
The company installed software that tracks how much AI code it generates. CTO Bart Butler said about 90% of October's code was written by AI. Yen told staff 2026 "will be the last year that we hire engineers."
After the story broke, Yen admitted "a communication failure on my part" and cancelled the re-interviews. Butler went further: "We fucked up, we know we fucked up." But one employee told 404 Media the things that would have been measured will now "become a component of the performance reviews."
So the exam is not gone. It moved into the appraisal.
My take: measuring volume is the mistake
I'm not against companies expecting AI fluency. I teach it. What backfires is measuring how much AI people use, without asking whether the work got better or stayed safe.
Track lines of AI code and you get more lines of AI code. You do not automatically get secure code. At a privacy company, that gap is the whole product. Proton's own users noticed: many said publicly they were cancelling paid plans.
And when people feel their job depends on looking AI-heavy, they stop raising concerns. 404 Media described exactly that: a culture where staff felt unable to voice security worries.
Why this matters in India right now
This is not a Swiss story. Korn Ferry's Workforce 2026 India findings say 70% of Indian employees report AI fluency is already incorporated into employee reviews, against 50% globally. The same survey says 72% feel they are doing more than one role.
The Economic Times put it well on October 30: frequent AI use "does not necessarily indicate better performance." Some Indian leaders already say the same thing. LinkedIn's Kumaresh Pattabiraman told the Times of India the company evaluates "outcomes and how they do things," not raw AI usage.
If your manager hasn't defined what "AI fluency" means in your review, assume it will be judged loosely. Bring your own evidence.
How to prepare for an AI fluency review
- Pick three real tasks. Not experiments. Work that shipped: a report, a test suite, a client deck, a reconciliation.
- Show before and after. Time taken, error rate, turnaround, customer response. One honest number beats ten vague claims.
- Show what you checked. What AI got wrong and how you caught it. This is the skill that separates fluency from dependence.
- Show where you said no. Client data you didn't paste in. Code you insisted on reviewing. That is judgment, and good managers reward it.
- Show it doesn't sound like AI. If your output reads like generic chatbot text, it signals low effort. I wrote a full guide on that: How to use AI without sounding like AI.
- Ask your manager one question. "How will AI fluency be assessed in my review?" If the answer is "usage," offer outcomes instead.
What I'd tell a manager
If you're designing these reviews, measure three things together: outcome, quality, safety. Reward people who flag risks. A metric that punishes the person asking for a pentest will eventually cost you more than it saves.
Proton's lesson is not "don't push AI." It is that pushing AI by counting usage, and through fear, breaks trust with both employees and customers.
Your appraisal will ask about AI. Answer with what you achieved.
If you want structured practice turning AI use into results you can show, see our programs at The Arthi AI Collective: https://thearthiaicollective.com/programs
Sources
- 404 Media, "Manifestos, Fear, and 'Loser Thinking': Inside Proton's AI Reckoning", 29 Oct 2026
- 404 Media, "Proton's CEO Cancels AI 'Re-Interviews'", 30 Oct 2026
- Times of India, "Swiss tech company Proton says 2026 is its last year of hiring engineers", 30 Oct 2026
- India Today, "Will it, won't it? Microsoft and Proton indicate AI will lead to job troubles", 1 Nov 2026
- Korn Ferry, Workforce 2026 India findings
- The Economic Times, "How should companies measure performance when AI does part of the work?", 30 Oct 2026
- Times of India, "AI fluency, outcomes key to appraisals"