90% of Its Code Is Already Written by AI: Proton Cancels Its Controversial Employee Re-Interviews After Backlash

Proton, the Swiss privacy company behind Proton Mail, has called off the internal "re-interviews" that would have put every employee's AI usage under scrutiny starting November 4, according to reporting by 404 Media. CEO Andy Yen canceled the sessions — which had been scheduled for November 4 to 11, with outcomes that could theoretically change roles or end jobs — and told staff there will be a reorganization but "no loss of jobs from AI."

The climb-down follows a 404 Media investigation published on October 29, based on interviews with eight current and one former employee plus internal Slack messages, meeting recordings and a memo from chief technology officer Bart Butler. Butler's three-page manifesto, posted to the company's Slack in late October, argued that engineers are no longer paid to write code: their job, he wrote, is to help large language models write better code faster, and most code at Proton should be reviewed by AI rather than humans.

The numbers attached to that push were striking. Butler said in a meeting that roughly 90 percent of Proton's code in October was written by AI, much of it with Anthropic's Claude. Yen set up a dedicated email address for staff to report anything blocking AI adoption so leadership could "ruthlessly smash" the obstacles, and the company installed software that tracks how much of its codebase is AI-generated. Yen also told employees that 2026 would be the last year Proton hires engineers in the traditional sense, with product-manager-style hires taking over from 2027.

The most inflammatory detail was an August Slack exchange in which a security employee asked for time to penetration-test new code. Yen replied that the request was, in the investigation's transcription, "bullshit and loser thinking." An internal document reviewed by 404 Media also showed Proton was not on track to hit its annual recurring revenue target by year-end, citing insufficient velocity — the business pressure behind the AI mandate.

Employees pushed back hard. Some feared that how enthusiastically they embraced AI could determine whether they kept their jobs, and several said a culture of fear made them hesitant to raise security concerns about AI-written code at a company whose core product is encrypted email. A Proton spokesperson did not dispute the reporting, saying some people were unhappy with the direction or misunderstood it, and acknowledging the company had "sometimes been too blunt or too slow" in explaining itself.

The cancellation was announced on November 1, though the retreat is partial: Yen said quarterly performance reviews will continue to measure AI adoption as a hard metric, so the underlying expectation has not changed — only the interview mechanism is gone. The episode quickly became a reference point in a wider industry argument. OpenAI engineer Peter Steinberger said Yen "gets it"; SpaceX principal engineer Lauren Tan wrote that "the time for mourning coding is long over"; Ruby creator David Heinemeier Hansson warned dissenting employees they were "in for a rough ride."

For a company that markets itself on privacy and user trust, Proton's stumble is a case study in how fast is too fast: the tools of surveillance-style productivity management — code provenance tracking, adoption quotas, re-interviews with job consequences — read very differently when applied to the people building the privacy suite rather than its customers. Whether the canceled re-interviews buy back any internal trust is another question the quarterly numbers will answer.