Proton says AI wrote 90% of its code as staff raised security concerns

Proton is pushing employees to use AI across software development, with its CTO saying that 90% of the company’s code for the month was AI-generated, according to meeting audio obtained by 404 Media. Multiple employees told the outlet that much of that code was written with Anthropic’s Claude.

The investigation is especially notable because Proton’s public reputation rests on privacy and security products such as Proton Mail, VPN, Drive and Pass. In one internal Slack exchange quoted by 404 Media, an employee asked for several days to penetration-test newly written code before release. CEO Andy Yen called the request “bullshit and loser thinking,” saying the company should focus on being faster while remaining secure.

An internal manifesto from CTO Bart Butler reportedly told engineers that their “old job is gone” and argued that Proton should move toward AI reviewing most code instead of humans. Yen also said the company would stop hiring traditional engineers and instead recruit people who can build products with AI.

Employees interviewed by 404 Media described a culture in which security concerns were difficult to raise. The outlet said it reviewed Slack messages, meeting recordings and internal documents and spoke with eight current employees and one former employee.

A Proton spokesperson did not dispute the reporting. The company said some employees were unhappy or anxious and acknowledged that leadership had “sometimes been too blunt or too slow” in explaining its direction.

The case puts a concrete number on the tension between faster AI-assisted development and the review standards expected from a company trusted with sensitive communications.