That said, I sincerely doubt this is the case. I imagine the phone acts as a proxy for the authentication, validating the fingerprint then sending some other form of authorization to Apple or using the fingerprint as input to a cryptographic algorothm.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3093498/ states that even professional forensics required independent verification to eliminate false positives.
The hashes, whatever they are, will not be "binary" in their nature. Matching against a range of visual characteristics requires to allow some level of fuzziness. Even assuming that near future improvements bring the false positive rate to half of that of the best forensic experts (to 0.05%), the law of large numbers guarantees that innocent people will be caught up in investigation dragnets. Just imagine the lives destroyed by these kinds of clerical errors.
The above is the same reason I'm against our national law enforcement getting access to the passport biometric databases. Even discounting the potential for abuse: once the police have a suspect with "matching" fingerprint available, they will have less incentive to find other ones.
No, it wouldn't send that response at all. That's called a client-side security control, and I'm sure you can think of why that's out of the question in any system.