Proton Mail’s parent company is internally pressing engineers to write and ship code with artificial intelligence as fast as possible, including over objections that new code should be pentested first, 404 Media reported Sunday after reviewing Slack messages, meeting recordings and an internal manifesto and speaking with eight current employees and one former employee¹.

The Swiss privacy company, known for encrypted email and related tools, has publicly argued that privacy firms should build AI so users are not left with large technology companies, including in an August interview Yen gave to WIRED that 404 Media cited¹. Internally, 404 Media said Yen’s tone shifted late in August toward using AI as much and as quickly as possible¹.

In the company’s security Slack channel at the end of August, an employee urged taking a few days to pentest newly written code before pushing it live, 404 Media reported. Yen replied that the delay was loser thinking, that it was possible to be fast and secure, and that the team should stop handcrafting all its code and use AI more, according to messages the outlet reviewed¹.

When another engineer asked how Proton would keep AI-generated code free of security issues and said security was a low priority, Yen did not answer the security point directly and told the channel that only losers thought speed meant going slower, 404 Media reported¹. The outlet said one employee stressed that engineers still try to build securely¹.

404 Media reported that Proton once banned internal AI use and later approved it for many tasks, including financial data, customer-facing Reddit and LinkedIn drafts, and writing code except when the source would immediately help a competitor or involved innovation or research, according to an internal document¹. In a meeting whose audio 404 Media obtained, CTO Bart Butler said 90 percent of Proton’s code in October was written by AI, much of it with Anthropic’s Claude, and that Yen also tells staff in Slack to use Claude¹.

Last month Butler posted a self-described AI manifesto to Slack that told engineers their old job was gone, said most code should eventually be reviewed by AI rather than humans, and called failing to transform an existential threat to Proton’s mission, 404 Media reported¹. Earlier this week Proton emailed staff that they would be re-interviewed about how they use AI and that the interviews were consequential, with workers telling 404 Media they understood jobs could be at stake¹.

An internal document obtained by 404 Media said Proton was not on track for its annual recurring revenue target because of insufficient velocity, that year-over-year growth lagged what the market considers good, and that by metrics that matter it was underperforming¹. The same document described an August outage as a catastrophic six-hour downtime, said moving to the United States or China was not an option, and proposed an elite fast team of 5 to 10 percent of staff with a 30 percent pay bump and weekend work, 404 Media reported¹. Yen has also described Proton as a 996 company in Slack, the outlet said¹.

A Proton spokesperson did not dispute 404 Media’s reporting or characterization¹. "It's clear that some people at Proton are unhappy with our direction, or misunderstand it, and many people are anxious," the spokesperson said, adding that the company had sometimes been too blunt and was committed to helping colleagues retrain¹. Whether the November re-interviews will lead to firings, and how AI-written code is tested in production, was not independently verified beyond 404 Media’s account¹.

Sources

1. 404 MediaManifestos, Fear, and ‘Loser Thinking’: Inside Proton’s AI Reckoning6:01 PM

Read the original at 404 Media ↗