Meta brings Private Processing to AI glasses for cloud-based assistant features

Meta says the system will let glasses handle more advanced AI features in the cloud without direct access to sensitive data.

Meta is bringing its Private Processing confidential-computing design to AI glasses, so heavier on-device assistant features can use cloud models and long-lived personal context without giving Meta direct access to the data. The setup relies on confidential virtual machines, TEEs, remote attestation, anonymous credentials, OHTTP relays, and encrypted storage, and Meta says it can support streaming transcription, contextual search, and long-term recall. Basic calls and messages still run on the glasses, while more advanced workloads move into Meta’s cloud boundary, which is meant to cover both CPUs and GPUs. The system is not yet available on the glasses, and Meta says it is announcing it early so security and privacy researchers, plus a third-party auditor, can review it; independent firms including NCC Group are also involved. Meta plans to give researchers CVM binaries, tools, and documentation as it prepares for more stateful, multimodal, and agentic AI features.