On-device AI: follow the data

Apple describes Apple Intelligence as using on-device processing where possible and Private Cloud Compute for some more complex requests. Google describes Private AI Compute as a protected cloud environment for features that need greater model capacity. These are vendor descriptions of architecture and privacy controls; they are not independent audits of every request path or enterprise configuration.

The hybrid reality

On-device processing can reduce data transfer for eligible tasks, but the privacy outcome depends on what data is processed, which route the system selects, applicable settings, account state, and any optional external service. A feature name alone does not prove that enterprise data remained on the device.

Enterprise checks

Document permitted features and request routes, test managed-device controls, review vendor retention and access commitments, and distinguish personal from managed accounts. For sensitive data, confirm the actual processing route and contractual terms rather than relying on a general privacy label.

Evidence boundary

The cited Apple and Google pages explain their own systems. Organisations should combine those materials with current technical documentation, contracts, device-management evidence, and a context-specific privacy assessment.