Proton reverses AI ban; 90% of code AI-written by October 2026

Proton built its reputation on privacy-first products including Proton Mail, Proton Drive, Proton Pass, Proton Meet and the AI service Lumo. According to 404 Media, the company had previously banned internal AI use before reversing course in August 2026. By October 2026, 90% of code was AI-generated, with Claude as the main tool. Internal measures described by 404 Media include a dedicated email address for reporting anything that obstructs AI adoption and software installed on employee PCs to track how much AI-generated code they produce.

The reporting also describes how dissent was handled. In late August 2026, an employee wrote on Slack that new code should be tested for several days before going to production; Andy Yen replied that this was nonsense and 'loser thinking,' said speed and safety can coexist, and told staff to stop thinking about how to slow down. Separately, after Reddit criticism of the May 2026 post-quantum encryption announcement — specifically that old emails keep the old encryption — Yen rebuked the responsible employees and told them to correct the facts on Reddit, saying perception matters more than truth.

Employees who spoke to 404 Media said colleagues hesitate to debate technical issues and fear being berated in front of others or losing their jobs, with one describing the CEO as fully in the grip of 'AI psychosis.' For a company whose selling point is trust in how it handles data, the reported gap between its public privacy stance and its internal pressure to ship AI-generated code is likely to be the point customers and regulators watch.