Apple only has one permission level for keyboards that send stuff over the network: full permission. There's no way to determine what exactly they send using apple's keyboard permissions API, and apple doesn't know what they send. They just know it's a keyboard that
can send anything you type over the network, and warn you accordingly.
You have to either trust that google isn't flat-out lying, or try to MITM and analyze the traffic using something like wireshark, which would be difficult assuming it's properly encrypted.