My favorite theory is this:
Apple wants to create a chilling effect on reverse engineering iOS. They're starting to catch regulatory attention and lost their recent suit against Corellium. By putting neural hashes in the encrypted OS image, they can accuse anyone who reverse engineers iOS of:
1. Being in cahoots with child abusers
2. Knowingly possessing CSAM
I would hope that most courts would see through that paper thin logic, but the idea would be simply to introduce enough doubt that most reverse engineers don't want to risk it.
NSLs are routinely issued to silence companies.
It was that Apple somehow gets negotiating leverage with a three letter agency.
So how does Apple get negotiating leverage in a context with a TLA when they can't say "they made us do it"?
What basis could there be for that? And I'd be even more skeptical if this originated from a single case of abuse.
How does Apple get leverage if Apple gets all the negative publicity, and can't even say "TLA made us do it"?
- The government is pressuring us to do things that don't fit with our "privacy is good" sales pitch
- Let's propose something, but in a ham-handed enough way that it gets a ton of public push back
- Now we can argue our point with the government in a way that has already has more public support and existing discourse/debate
Somewhat risky though, since it hurt their public image.