Proton's AI Pace Shows: Security Promises Require Evidence of Leadership
What this is about
The Geneva-based company Proton is pursuing an aggressive AI strategy. Research published on 29 October 2026 by 404 Media is based on conversations with eight current and one former employee, as well as Slack chats, meeting recordings and an internal AI manifesto. According to the report, CEO Andy Yen pushed for code to be written increasingly by AI rather than by hand and rejected a request for additional security testing as “bullshit and loser thinking”. Proton did not dispute the authenticity of the quoted chats when responding to 404 Media. Yen apologised publicly for hurtful remarks directed at employees who had helped build the company.
Why this matters for the board and executive management
Proton sells trust in secure communication. According to a reviewed meeting recording, AI wrote around 90 per cent of the code in October, whilst Yen dismissed a request for penetration testing before deploying new code to production as “loser thinking”. This touches on overall management and overall supervision under Art. 716a OR, since it must be clarified whether the pace has changed the review and approval processes for security-critical code. Additionally, Proton had planned a new survey on AI usage, which was also linked to job security; this was withdrawn on 30 October 2026. The board must be able to rely on the executive management demonstrating that it can substantiate the actual handling of these processes.
Questions for executive management
- Which review and approval steps does security-critical, AI-generated code currently go through, and has their depth changed since the acceleration began?
- By what criteria will employees be assessed in future, now that the announced survey has been withdrawn?
Sources
- Manifestos, Fear, and 'Loser Thinking': Inside Proton's AI Reckoning (Primary source, 404 Media, 29.10.2026)
- Proton gerät mit aggressivem KI-Kurs in die Kritik (SwissCybersecurity.net, 10.11.2026)
- Proton gerät mit aggressivem KI-Kurs in die Kritik (Netzwoche, 10.11.2026)