By Sloane, PR and Marketing, HUMAN HIVE
Proton told staff their AI use would be judged in "consequential" re-interviews, then cancelled them four days later after leaked messages went public and drew a backlash. The tools weren't the problem. The rollout put pressure ahead of practice, and people pushed back.
The 60-second version
Proton planned AI "re-interviews" for staff from Nov 4 to 11, tied in an internal email to "continued employment."
After leaked messages went public, the CEO cancelled them on Oct 30 and the company apologized publicly.
The lesson for every leader: say what AI is for, train the habit first, and never make fear your adoption plan.
What happened
On October 26, 2026, Proton, the Swiss company behind Proton Mail and Proton VPN, told staff it would be "re-interviewing" them about how they use AI. A leaked email seen by 404 Media read: "These re-interviews are consequential," and tied the process to "continued employment." The Next Web reported that the sessions were set for November 4 to 11, run by interviewers from outside each employee's team who must "demonstrate strong AI capability."
Then came the messages. According to 404 Media, when an employee asked in late August for time to security-test new code, CEO Andy Yen called the idea "loser thinking" and wrote: "Stop thinking always of how to be slower, and focus more on how to be faster." CTO Bart Butler said in a meeting that AI wrote 90% of Proton's code in October.
On Friday, October 30, Yen cancelled the re-interviews at an all-hands meeting. He acknowledged a communication failure on his part and said he would keep pushing on AI. A Proton spokesperson told The Next Web: "In our determination to move fast, we have sometimes been too blunt or too slow in explaining ourselves to our people." Proton says it is not cutting jobs and is still hiring. In a public statement, the company apologized and promised to publish how its code and security reviews work and where AI is and isn't allowed.
On November 1, Butler published his internal memo, "Proton's AI manifesto for software engineering." One line in it is the key to the whole story: "Every engineer is now a team leader, a manager, and your direct reports are the models and agents you use."
Why it happened: the human side
Pushing hard on AI is a reasonable bet. What turned it into a fail was the order of operations. We call the pattern mandate before method.
The threat arrived before the training. People heard "your job depends on this" before they heard "here is how we'll help you get good at this." Fear doesn't speed adoption up. People hide what they don't know.
Speed was set against care. When a leader calls a request for testing "loser thinking," everyone else learns to stop raising risks. That's the most expensive habit a team can learn, and it shows up long after the tool rollout.
The new job was named but not taught. If every engineer is now a manager of AI agents, that's a leadership skill. Leadership skills are trained and practiced. They don't arrive in a memo.
The data backs this up. Gallup found that "Employees whose managers actively support AI use are nearly twice as likely to use it frequently." As of May 2026, only 36% of employees in AI-integrating organizations strongly agreed that their manager supports their team's use of AI.
What Mitch has taught on this
In his video "The three things you need to know in the age of AI," Mitch makes the point Proton learned the hard way: "the cost of knowledge is now zero and that knowledge is not wisdom." A memo is knowledge. Putting it into practice takes wisdom.
He goes further: "AI can only give you information." Even if it hands you steps for a training policy, "it can't actually implement the training policy." And on leadership: "If it gives you a blueprint for leadership and you've never been a leader and you try to implement the blueprint, you will fail because you don't have the experience."
That's the gap Proton found. It wrote a blueprint that made every engineer a manager, then tested them like it was a pop quiz.
Best practices: five moves before your next AI push
Write the "why" in one page before any mandate. The executive sponsor writes what AI is for here, where it's allowed, where it isn't, and what is not at stake (jobs, bonuses). Proton promised to publish this after the backlash. Write yours before.
Train before you test. L&D runs short, role-based practice sessions on real tasks first. Any assessment comes after people have had a fair chance to learn, and it's framed as coaching, not a trial.
Protect the person who raises a risk. Managers thank the first person who says "should we test this?" in public, then answer the question. The rule for leaders: you can say "not now," never "loser."
Make managers go first. Each manager shows the team one real way they use AI each week, including a time it got something wrong. Gallup's data says manager support is where frequent use starts.
Measure outcomes, not AI percentages. "90% of code written by AI" says nothing about whether customers are safer or happier. Track the result the work exists for: quality, speed to customer, errors caught.
What's driving people crazy this week
AI mandates that feel like threats. The Proton re-interview email, as 404 Media reported it, linked AI use to "continued employment" before most people knew what "good" would look like. Employees read it as a test of their job, not their tools. The human fix: separate learning from judging. First you train people, then you coach them, and you measure only after both.
What this means for the next 1 to 3 years
Here's our view. First, "AI fluency" will move into performance reviews at many companies, and the ones that win will define it as judgment and results, not usage counts. Second, public AI policies become part of the brand: Proton's customers cared how its code was made, not just what it does, and buyers in every industry will start asking the same question of their vendors. Third, the title "manager" spreads to people who manage agents, which means leadership training stops being a perk for the few and becomes basic training for everyone.
How we evolve
The human skill to build is coaching: helping someone get better without making them afraid. It matters more as AI gets better, because more of everyone's work becomes directing, checking and correcting, which is exactly what good coaches do with people. It's the heart of Mitch's ADAPT! message too: adaptability is the skill, and people build it by trying, not by being afraid of getting it wrong. One daily practice: ask one person on your team, "What did AI get wrong for you this week, and how did you catch it?" Then just listen.
Try this today (under 30 minutes)
Write your team's one-page AI "why" using the template below, and share it as a draft.
At your next team meeting, ask: "What would you need to feel confident using AI on your real work?" Write down every answer.
Pick one risk someone has raised about AI in the last month, and reply to them today with a thank-you and a decision.
Free tool: the one-page AI "why," ready to copy.
1. Here's what we want AI to help us do: [one outcome].
2. Here's where it's welcome: [tasks]. Here's where it isn't: [tasks, data].
3. Here's the training and time you'll get: [sessions, hours].
4. Here's how we'll know it's working: [result, not usage].
5. Here's what is NOT at stake while you learn: [say it plainly].
FAQ
Did Proton fire anyone over AI use?
Proton says it isn't cutting jobs and is still hiring. It cancelled the planned re-interviews on October 30, 2026.
Is it wrong to expect employees to use AI?
No. Expecting it is fine. The fail is expecting it before people have training, clear rules and a manager who uses it too.
What's the first step for a leader rolling out AI?
Write a one-page "why" that says what AI is for, where it isn't allowed and what isn't at stake, then train before you measure.
The failure was a decision, not a tool. Test how your team decides before it costs you.