Converge Digest

Apple powers Private Cloud Compute with its own servers

At its #WWDC24 event in Cupertino, Apple announced Private Cloud Compute, its in-house cloud intelligence system designed specifically for private AI processing.

Apple Private Cloud Compute Foundation Models

The idea is to extend Apple Intelligence from the user’s device to the cloud while preserving security and privacy.  Apple says this will enable user to access even larger server-based models than could be achieved on-device, while ensuring that data stays within its Private Cloud Compute infrastructure. The company promises end-to-end encryptions as its best defense. The user data also must not be retained on servers, including via logging or for debugging. The data must also not be accessible by Apple and this means that the PCC infrastructure must not contain privileged interfaces. In addition, Apple says its guarantees must be enforceable and open to third-party verification. Here is a blog posted today on the topic:

Apple’s Craig Federighi said Private Cloud Compute will be powered by custom servers built with Apple silicon.  He also noted that user data would not be used to train Apple’s AI models.

For each AI request from a user device, an encrypted tunnel is established to a specific server cluster. Only that cluster has the key for the encrypted tunnel. The data is not retained after the response is returned. The session is not accessible to Apple.

https://machinelearning.apple.com/research/introducing-apple-foundation-models

https://security.apple.com/blog/private-cloud-compute

Apple-owned or operated Data Centers

United States:

China (in partnership with local providers):

Europe:

In addition, Apple confirmed that iOS18 will have SMS-via-satellite capabilities.

Exit mobile version