Deepinder Goyal's new project promises E2E-encrypted AI chats on self-hosted open-weight models — but where do chats decrypt?
rajupp · x · 2026-10-22
Deepinder Goyal's new venture promises end-to-end encrypted AI chats that even his company Eternal can't read, running self-hosted open-weight models — a contrast to ChatGPT's Locked Chats, which showed a PIN isn't encryption. But tech commentator rajupp raises the key question: if the model runs on their servers, where exactly do chats get decrypted? He tagged Goyal directly awaiting an answer.
Related event: E2E-encrypted AI chat sparks debate as Meta open-sources Private Processing(2 posts)→
More from Safety
- Scoble: your AI systems will be attacked — spend tokens on a security audit now — IgorCarron · 2026-10-22
- AI voices clash over EU AI Act: regulated AI 'better for everyone' vs compliance-run future — DionysianAgent · 2026-10-22
- Cryptographer Matthew Green jokes AI's bug-finding speed means 'it'll all be over in six months' — matthew_d_green · 2026-10-22
- Capcom confirms cyberattack exposing 11.77 million CAPCOM ID accounts — CtrlAltDwayne · 2026-10-22
- Columbia paper: reward misspecification breaks CoT monitoring; new training lifts detection 18%→85% — burny_tech · 2026-10-22
- Researcher: eMudhra's ISO27001 auditor's '3rd floor' office doesn't exist — kingslyj · 2026-10-22