AI AI Toolkit
AI Newsindustry

CEO Andy Yen 强推 AI 引发恐惧文化

404 Media(RSS)2026-10-29T14:21:49.000Z

Key Highlights

On October 29, 404 Media published an investigation into the encrypted-email provider Proton's internal AI reckoning. Based on interviews with eight current and one former employee, plus Slack logs, meeting recordings, and a CTO-written AI manifesto, the report concludes that Proton, a company built on privacy and security, is pushing AI internally at a startling speed while many employees dare not voice safety and quality concerns. The piece is a rare look behind the brand.

What Happened

The most striking detail is from late August, when CEO Andy Yen snapped at an employee in the security Slack channel who suggested pentesting new code for a few days before shipping. Yen wrote that was bullshit and loser thinking, insisted it is of course possible to be fast and secure, reiterated that only losers think like that, and ordered the team to stop lovingly handcrafting all our code and use AI more. He also set up a dedicated email to collect anything blocking AI adoption so leadership could ruthlessly smash it, installed software tracking employees' AI-code output, and said this is the last year Proton hires engineers, after which it will only hire people who build with AI.

Technical Details

The AI push peaked this week when 404 Media reported Proton would re-interview staff to assess AI usage, implying the interviews, and willingness to embrace AI, could decide who keeps their job. An internal document obtained by 404 Media says that due to insufficient velocity Proton is off track for its annual-recurring-revenue target, year-over-year growth is far away from what the market today considers good, and by all metrics that matter, we are an underperforming company. The document reads like a mandate for speed at any cost.

Versus Competitors

Ironically, Proton projects a cautious, privacy-protecting image publicly while sidelining the security-versus-AI-speed conflict internally. One source said flatly that the CEO is in full AI psychosis. This contrasts sharply with privacy companies that treat AI-assisted coding more carefully and keep security review in the loop rather than treating it as an obstacle to be smashed.

Industry Impact and Use Cases

For teams that rely on Proton's privacy products, or are themselves forcing AI adoption internally, this investigation is a warning bell. Driving AI speed through a culture of fear can come at the cost of security and trust. When a company that sells encryption starts mass-generating code with AI while suppressing security objections, the industry's governance gap around AI-assisted development deserves a fresh look.

Practical Notes

The lesson for engineering leaders is that velocity metrics tied to AI-code output are a double-edged sword. Tracking how much code is AI-generated can surface adoption, but when it becomes a proxy for performance reviews, it incentivizes shipping unreviewed code. Pair any such metric with mandatory security gates and independent review so speed does not quietly become risk.

What to Watch

Watch whether Proton's security track record holds up over the next few quarters. If breach or quality incidents surface, this reporting will be cited as a preview. The broader trend, measuring engineers by AI throughput, is spreading across the industry and warrants scrutiny before it becomes the default management idiom.

Takeaway

Proton's internal AI reckoning is a case study in how a privacy-first brand can fracture when growth pressure meets uncritical AI enthusiasm, and it should be required reading for any leader mandating AI adoption under deadline.

Background

Proton has long marketed itself as the privacy-first alternative to Gmail, so an internal culture that punishes security caution is a story with real reputational stakes that go beyond one company. 404 Media's sourcing, eight current and one former employee plus internal logs, makes this more than a rumor, and it raises a question other privacy vendors should ask themselves about how they handle AI-speed pressure.

Where to Watch Next

Follow whether any of the named practices change after publication, because a response or a revised AI-code policy becomes the second half of the story. The episode is also a useful case for any company mandating AI adoption under performance pressure, showing how quickly a stated value like security can be reframed as an obstacle to be bypassed.

Closing Thoughts

The lesson is that culture, not tooling, decides whether AI adoption stays safe, and a leader who equates caution with losing will eventually ship the bug that the caution was meant to prevent.