Proton's AI reckoning: 'loser thinking' and re-interviews

What the investigation found

404 Media published an investigation on October 29 into how Proton, the Swiss maker of Proton Mail, Proton Drive, an encrypted authenticator and a password manager, launched in 2014 after the Snowden revelations, and other privacy products, is pushing artificial intelligence through its engineering culture. The report is based on interviews with eight current and one former employee, plus Slack messages, meeting recordings and an internal document described as the chief technology officer's AI manifesto.

The central figure is CEO Andy Yen, who publicly presents a measured view of AI while internally pushing staff to adopt it quickly. The reporting describes a gap between Proton's public reputation as a careful protector of privacy and an internal push that employees say leaves little room to raise security concerns.

Proton did not dispute the reporting. A spokesperson told 404 Media that some people at the company are unhappy with the direction or misunderstand it, and that many are anxious. The spokesperson said Proton had sometimes been too blunt or too slow in explaining itself.

The security argument and the 'loser thinking' reply

One exchange began when an employee in a security Slack channel asked the company to take a few days to test newly written code before pushing it live. Yen replied that the idea was "bullshit and loser thinking" and that it was possible to be fast and secure. He told the channel to stop thinking about how to be slower and focus on being faster, then added that "only losers think like that."

The message ended with a broader instruction: stop handcrafting code and use AI more. Another engineer pushed back, saying AI would not solve security problems because security had a low priority at the company. According to 404 Media, Yen did not answer the substance and instead criticised another employee in front of the channel.

Employees told the outlet that this was not an isolated tone. One described a culture where people "just shut up and get on with things." Another said there was "100%" a culture of fear, especially in recent months. A third said Yen can be hostile in public chats when he disagrees with a view. Employees told the outlet they were hesitant to criticise Yen's views or raise engineering problems.

The CTO's manifesto and 90% AI-written code

Proton's chief technology officer, Bart Butler, posted what 404 Media describes as a self-described AI manifesto in Slack on a Friday afternoon. It told engineers that their old job was gone and set out three changes: engineers should help large language models write better code faster, most code should eventually be reviewed by AI rather than humans, and the organisation should commit to the transformation.

Butler said in a meeting that 90 percent of code at Proton in October was written by AI, according to audio obtained by 404 Media. Employees said much of it was produced with Anthropic's Claude, which Yen also encouraged staff to use.

The manifesto was not universally welcomed. One employee replied that scaring people on a Friday was unnecessary. Another told 404 Media that the document exaggerated what language models can do today and did not address the problems that come with handing a codebase to AI, including long-term maintenance. One employee called the manifesto "ethically bankrupt."

The re-interviews and what they mean for jobs

This week Proton told employees that they would be re-interviewed about how they use AI in their work. The email said the interviews were "consequential," a word employees read as a warning that the outcome could affect their roles. The interviews are scheduled for November 4 to 11 and will be conducted by senior staff chosen for strong AI capability, according to heise online's coverage of the same reporting.

Yen told one senior employee who questioned the process that leadership already knew roughly who was behind and that the company would not bother with the process if it were giving up on them. The employee replied that the messaging was fear-based and that a fear-based culture was "a cancer" for engineering. Yen answered that if the employee did not trust him when he said the decision was not financially driven, "this might not be the place for you."

Proton has said it is not cutting jobs and is not using AI as a pretext for cuts, according to heise. The tension in the reporting is not a formal layoff announcement but a process that employees believe could reshape who stays.

Not everyone will be interviewed, according to heise, and Proton says it recognises the uncertainty and will share individual outcomes as soon as possible. The company had about 500 employees in its latest 2024 figures.

The business pressure behind the push

The investigation links the AI drive to performance pressure. One internal document said that "due to insufficient velocity" Proton was not on track to meet its annual recurring revenue target by the end of the year, that year-over-year growth was far from what the market considers good, and that "by all metrics that matter, we're an underperforming company."

Yen, a former CERN particle physicist, has also set up an email address for employees to report anything blocking AI adoption, so leadership can remove obstacles to what he calls AI velocity, and Proton installed software that tracks how much AI-generated code the company produces. He said this would be the last year Proton hires engineers, with future hires focused on building with AI.

That sits awkwardly with Proton's public positioning. The company, whose products include Proton VPN, Drive, Pass and Calendar, has built its brand on encryption and privacy, and its AI chatbot Lumo is marketed as not logging conversations or training on user data. In an August interview with WIRED, Yen framed privacy-first AI as a way to avoid ceding the future to companies that do not have users' interests at heart.

What is not known

404 Media published images of the version of the manifesto it obtained, and Proton has not published the document itself. The version it published was retyped by the outlet. The exact number of employees who will be re-interviewed, and what happens to those who decline or perform poorly, is also unconfirmed.

Proton's statement does not address the specific incidents described by employees. Whether the AI push improves security and product quality, or creates the risks employees describe, will only become clear over time. No public timeline has been given for the re-interviews beyond that window.

The company has not said whether the output of that tracking software is shared with employees, which is one of the practical questions the internal messages leave open. Related coverage includes Nvidia antitrust questions, Anthropic cyber verification tiers, and Industry. The site also covered UK ICO guidance on agentic AI.