Proton 爭議的核心並非 AI，而是管理層對安全測試的態度

I’ve been paying for Proton for many years, and it’s what I recommend when someone asks how to get off Gmail. This week 404 Media published an investigation into how Proton’s leadership is pushing AI inside the company. The detail that made the rounds came from the security Slack channel: an engineer asked for a few days to pentest new code before shipping it, and CEO Andy Yen answered that it was “bullshit and loser thinking”, then told the team to stop lovingly handcrafting code and use AI more.

For the first time, I’m not sure I’d still recommend them.

A lot of the reactions treat this as an AI story, as if any AI-written code poisons the well, and I think that misses the point. Engineers using AI to write code is the new norm. It’s fine as long as there are guardrails around it: someone reviews what the model produced, it gets tested, and someone is accountable for what ships.

What the reporting describes is a CEO telling an engineer, in front of colleagues, that raising a security concern was “loser thinking”. Staff were told their AI usage could decide whether they kept their jobs, and when that plan was dropped after it leaked, Yen said it had already served its purpose. None of this is about AI. Plenty of companies are rolling out coding agents right now without humiliating anyone, so if Proton can’t, the problem sits with its leadership.

The tradeoff Proton can’t make

For most companies, trading some caution for speed is a reasonable call. Proton is a privacy company, though, and what its users pay for is the confidence that the people building it are careful. Shipping slowly and carefully is part of what they’re selling. A bad release from a B2B vendor can be catastrophic, but for a company like Proton, a bug that exposes users’ mail would be fatal.

One of the employees defending the company on Reddit added some context that was meant to help: Yen was fine with pentesting, as long as it didn’t slow down the pipeline.

This is one of a few employees pretending to courageously dissent, but in reality agreeing with the official corporate line. It reads like astroturfing and a clumsy attempt at crisis management. No employee who shared on Reddit disagrees with how things were handled or with where the company is going.

Beyond the clumsiness, it shows a leadership that now treats the constraints that made Proton worth paying for as the thing slowing it down.

I’d like to keep recommending Proton, but that depends on the engineer asking for a few more days of testing still winning the argument against a leadership obsessed with velocity and arbitrary release dates.