Apple officially called the use of larger basic models for complex data reasoning advanced capabilities, decided to create a private cloud computing, is a breakthrough cloud intelligence system designed to private AI processing. In short, the business will be deployed to the cloud with security and privacy.

It should be known that Apple has always insisted on data processing directly on the device, only in the form of data on the user's device, which can avoid centralized attack points. Apple City will strengthen end-to-end encryption after going to the cloud. For cloud servers that are not suitable for use, user data is processed in a temporary manner or with irrelevant random identifiers to mask user identity.

Powerful AI hardware in the data center can fulfill user requests with large, complex machine learning models, but there are some challenges such as:

Cloud AI security and privacy guarantees are not easy to verify and enforce. AI services such as Cloud claim not to record user data, but security researchers can't verify a mine, and server vendors can't enforce that promise for long. Such as new versions of AI may introduce additional logging, inadvertently recording sensitive user data but the technical work can not be serious. A peripheral load balancer that terminates TLS may log thousands of user requests in a troubleshooting session.

Cloud AI provides runtime that is not transparent. For AI environments, the tight restrictions on privileged access are also a challenge. Cloud AI services are complex and expensive to run at scale. Runtime performance and other operational metrics are also constantly monitored and investigated by vendor site reliability engineers and other administrators. These people usually have high privilege to access services.

Apple has built the cloud with the following solutions: stateless computing, enforceability guarantees, unprivileged access, unlocability, and verifiable transparency.

Perform stateless calculations on personal user data. Private cloud computing must receive individual user data exclusively for user requests and is not provided to anyone other than the user. Even employees cannot access it during active processing. This means that users' personal data will not leave a trace in the pcc system. Unprivileged operation, non-targeted, verifiable cloud service transparency. Apple introduces private cloud computing nodes. Uses custom server hardware, uses secure IOS security technology, and has a set of custom cloud extensions. Let users firmly believe that the private cloud is the leading large-scale AI computing security architecture, and the application of private cloud continues Apple's strong commitment to user privacy.