It wasn't AI that flipped Proton's charge
Raed Chammam's blog post responds to a 404 Media investigation published October 29, 2026, which described pressure inside Proton, the Swiss company behind Proton Mail, to build code with AI. The central episode dates to late August. An engineer in the security Slack channel asked for a few days of penetration testing on new code before release. CEO Andy Yen replied that this was “bullshit and loser thinking” and told staff to stop “lovingly handcrafting” their code and use AI more.
The reporting describes more pressure than that single exchange. Yen reportedly wrote in late August that this would be the last year Proton hires engineers. CTO Bart Butler posted a September manifesto saying engineers' old job was gone, and in a meeting said AI wrote 90 percent of the company's code in October, much of it with Claude, according to employees. Proton also told staff that AI-use re-interviews were consequential, and it installed software on staff computers that tracks how much code AI generates.
Proton did not dispute the reporting, according to 404 Media, and a spokesperson said some staff were unhappy or anxious. The report draws on interviews with eight current employees and one former employee. Reports differ on the re-interviews: TheNextWeb said they would begin November 4, while Fast Company said they had been canceled after the investigation. Chammam's position is that the core problem is leadership, not AI.
Why it matters
For AI companies, the backlash centers on how security review was handled, not on the tools engineers used. Chammam argues that AI-written code is acceptable with proper guardrails, and for a privacy company those guardrails are the product. Cybernews reports that some Proton users said they canceled subscriptions after the coverage, and some called for Yen to resign. Customer trust is the asset a privacy vendor sells, so reputational damage reaches revenue quickly.
For investors and operators, the talent risk is just as concrete. The reporting says employees felt unable to raise security concerns about AI-driven products for fear of being berated or losing their jobs. A stated plan to stop hiring engineers and rely on product managers who build with AI changes the talent base that security-sensitive products depend on. Diligence on AI-heavy engineering teams should ask who can block a release, not only how much code is generated.
Key points
- 1The disputed Slack exchange took place in late August in Proton's security channel.
- 2Proton's CTO reportedly said AI wrote 90 percent of the company's code in October.
- 3Proton's email called AI-use re-interviews consequential, and reports conflict on whether they were canceled.
- 4404 Media's October 29 report cites eight current employees and one former employee as sources.
- 5
