When staff stop raising AI concerns: what Proton's cancelled re-interviews tell leaders
· 5 min read
Proton cancelled AI re-interviews after reports staff felt unable to raise concerns; new research shows the same quiet gap between managers and teams elsewhere.
In brief
Proton, the Swiss company behind Proton Mail, cancelled plans to interview staff again about their AI use after 404 Media reported on its aggressive AI push and on employees who felt unable to voice security concerns. Its chief executive admitted a communication failure but said the push would continue. The episode is unusually public, yet the pattern behind it is common. New vendor research from Perceptyx suggests that most managers feel ready to lead AI change, while far fewer employees say their manager actually helps them adapt. Executives are also much more comfortable raising concerns about AI tools than the people who use them every day. This article looks at what both sources say, why a strong push for adoption can quietly silence the people closest to the problems, and what leaders can do about it. It points to practical tools from the AI Enablement Toolkit for Leaders, including a change listening guide, an audit of what your measures really reward, and a short regular review that asks what people are telling you that you may not want to hear. It also explains where manager training helps, and it is clear about the limits: the toolkit is a leadership method, not a security review, and not legal or HR advice.
A re-interview for the job you already do
Late last week, 404 Media published an investigation into Proton, the Swiss company behind Proton Mail and Proton VPN. Based on employee interviews, internal messages and meeting audio, it described a chief executive pushing AI hard and employees who felt unable to voice security concerns. Staff were reportedly due to be "re-interviewed" about how they use AI at work, with the clear implication, according to the employees 404 Media spoke to, that their jobs could depend on the answer.
On Friday 30 October, chief executive Andy Yen cancelled the re-interviews at an all-hands meeting. He said: "I want to acknowledge myself, personally, that there was a communication failure on my part". He also said there would be "no loss of jobs from AI", while making clear the direction had not changed: "Going forward, I'm still going to push." On 31 October Proton posted on X that it was not re-interviewing staff, that there were no job losses due to AI re-interviews, and that it would publish its internal AI policies. It also denied installing tracking software to measure AI-generated code.
Reactions inside the company were mixed. One employee told 404 Media: "As an employee, the re-interviews weren't cancelled until this became public despite concerns raised by many of us." Another said the things that would have been measured in the re-interviews would now become part of performance reviews.
Proton's story is unusually public. What sits underneath it is common.
What the research says about the quiet middle
On the same day, Perceptyx, a company that sells employee listening and survey software, published a review of employee sentiment on AI drawn from its benchmark database and a 2026 panel of employees in North America and Europe. Treat it as vendor research. Its findings line up uncomfortably well with the Proton story.
Two thirds of employees (67%) believe AI will help their organisation stay competitive. But only 33% say they feel well prepared to use AI tools in their daily work, and just 31% say their organisation has a clear plan for increasing adoption. 56% trust their organisation to use AI responsibly.
The gap that matters most for leaders is between managers and the people they manage. 60% of managers say they feel ready to lead AI-augmented work, yet only 44% of employees say their manager helps their team adapt to AI-related change. In Perceptyx's 2025 research, 84% of executives felt comfortable raising concerns about generative AI tools, compared with 59% of individual contributors. As the report puts it, individual contributors "tend to be closest to the practical problems a new tool creates, so when they hold back, leaders lose an early source of information."
Where managers do help their teams adapt, 82% of employees trust their organisation to use AI responsibly, against 56% across the panel. Perceptyx is careful to say these are relationships, not proof of cause.
What it means for leaders
Most organisations will never face an investigative story about their AI programme, which is why the lesson is easy to miss. The cost of not hearing concerns usually arrives quietly: a security shortcut nobody challenged, a tool used for show because usage is what gets counted, a capable person who stops saying "this doesn't look right" because the last person who said it was mocked.
Ian Hughes describes the mechanism in What Stays the Same:
"The objection you cannot hear is the one running your organisation, whether you like it or not."
A push for adoption creates its own silence. When AI use becomes a test of loyalty or a line in a performance review, people learn to show the behaviour that is measured and keep their doubts to themselves. That is not resistance. It is a sensible response to the signals they are being sent.
Where the AI Enablement Toolkit for Leaders helps
Esm's toolkit is a leadership method, not software, and several of its 26 tools speak directly to this problem.
- Tool 3, the change listening guide, gives a leader seven moves for a one-to-one or small team conversation, from "What part of this change feels most useful, and what part concerns you?" to "What do you think we could lose if we get this wrong?" Its discipline is blunt: if the answer changes the plan, that is evidence the listening worked.
- Tool 13, the incentive and behaviour audit, compares what you say you want with what actually gets praised, promoted and measured. If AI use is about to enter performance reviews, this is the place to test whether you are rewarding use or useful outcomes.
- Tool 24, the ten-minute leader check-in, builds one question into every review: "What are people telling us that we may not want to hear?"
- Tool 25 lists "Culture traded for compliance" as an anti-pattern, where people obey while concerns and useful dissent go underground.
The Perceptyx manager gap is also where training earns its keep. Managers absorb most of the change conversations, often while their own work is changing too. Esm's training and enablement work focuses on that leadership capability: helping managers hold honest conversations about AI, spot a stalled adopter, and turn what they hear into a decision rather than a communication task.
Honest limits
The toolkit is a management control, not a technical fix. It will not review AI-generated code for security flaws or replace a proper security process. It is not HR or legal advice on performance management or consultation duties; check the rules that apply to you rather than relying on memory. The Proton account rests on 404 Media's reporting and Proton's own statements, and the company did not respond to 404 Media's request for comment. The Perceptyx figures come from a vendor, describe associations rather than causes, and mix 2025 and 2026 data.
What the toolkit can do is make listening a habit with a name and an owner, before the concerns find a less friendly audience.
Tags: AI adoption, employee voice, change leadership, AI governance, managers
Sources
Join the Exchange free
Join free to unlock the complete knowledge library.
Founder membership is free for life
More from Esm
How should IT handle old SharePoint data when every department wants to keep everything?
4 min read
AI skills gap: why two thirds of UK firms using AI say they lack the know-how
5 min read
How do I find everyday AI uses for staff outside IT?
5 min read
Experience debt: what happens to judgement when AI takes the tasks people learned from
5 min read
How should IT teams handle emergency patches for critical vulnerabilities?
5 min read
When the human in the loop becomes a rubber stamp: what UK regulators found about AI checks
5 min read
Related articles
How should IT handle old SharePoint data when every department wants to keep everything?
Treat old data as a business decision IT runs as a service: one policy owner, agreed retention periods, clear archive tiers and automated rules.
AI skills gap: why two thirds of UK firms using AI say they lack the know-how
The UK government's AI adoption survey 2026 finds adoption has doubled but skills are the biggest barrier, even for firms already using AI. What it means for leaders and which toolkit tools help.
How do I find everyday AI uses for staff outside IT?
Start with the work your teams already do, not a list of AI tricks: observe, set data rules, run small timed trials with an owner, and publish what works as knowledge.
Experience debt: what happens to judgement when AI takes the tasks people learned from
New research finds most people say skills are slipping as AI takes on routine tasks, while time to learn is scarce. What that means for leaders, and which toolkit tools help rebuild the learning route.
How should IT teams handle emergency patches for critical vulnerabilities?
Treat an emergency patch as an emergency change: triage the risk fast, agree the fast path in advance, test as far as the clock allows and keep a remediation plan.
When the human in the loop becomes a rubber stamp: what UK regulators found about AI checks
The Bank of England and FCA find most firms approve AI before launch but far fewer keep watching it, and the FCA warns human review can become a tick-box exercise. What leaders should check, and which toolkit tools help.