It smells like data mining to me; instead of building their own safe browsing service, why not just defer to Google's services? I'm sure Apple pays Google for the service, and probably on a per request basis, so proxying and caching data will reduce their costs... In theory, I'm sure there's contractual obligations in place.

(from personal experience, we used a service in a project once to fetch stock market data. It was a good faith agreement, we had to maintain our own tallies of how often a stock price was fetched, tell them, and we'd be charged accordingly. This made sense because from them we just got a 20 mbit inbound pipeline of stock updates, which we fed into a Gemfire (which is now apparently called "VMware Tanzu™ GemFire®") cluster and exposed to end-users via a simple REST API. (It was a bank, of course we couldn't change firewall rules to allow websockets or whatever was available at the time).