Google DeepMind announces secure server-side memory for Private AI Compute
DeepMind detailed how Private AI Compute will use persistent, encrypted server-side memory with device-held cryptographic keys and hardware-enforced secure enclaves, plus public verification materials and audit results.
Announced a persistent server-side memory layer: encrypted per-user storage decrypted in hardware-enforced secure enclaves, with cryptographic keys held on personal devices; published a tamper-proof server software record and independent audit results.
Privacy teams and enterprises can verify server software authenticity and audit protections before sending personal data, supporting assessment of trust for cloud-assisted AI services.
Maybe — Review the updated Private AI Compute Technical Brief and verification materials.