Let’s Encrypt AI proposes confidential inference by 2027

The author argues that AI chats and inference should be protected by verifiable confidential computing using hardware-rooted trusted execution environments by the end of 2027. The proposal calls for encrypting prompts, responses, and context across the entire route from client to hardware, enabling remote attestation, and making verification a standard condition of service. The goal is to shift the market so that users retain ownership of their data and private work stays outside model training pipelines by default. Open article →