Without arguing the merit of your personal position, "this isn't Minority Report" is a terrible argument. Minority Report is about pre-crime. If someone has CSAM on their phone (and it's not a false positive), they are already actively committing a crime.
Ridiculous. So if I use malware to insert CSAM on your iPhone, you're "actively" committing a crime? CSA is a crime, but a person suspected of that must be properly tried through the judicial process. Private entities can also monitor what passes through their premises, but they have no business planting continuous surveillance functionality on your premises without a court warrant.
Everyone knows that the worst CSA happen in lawless countries and regions. This is mere pretext to build generalised surveillance infrastructure, or more importantly, normalise the concept of being continuously surveilled by faceless entities. That is the big push society is heading generally towards. It started with CCTV and will end with brain implants, unless society decided enough is enough at some point instead of endless hair-splitting.
The law still views willful possession as a crime, otherwise we wouldn't be here. I agree that does open up the possibility that you could frame somebody by planting CSAM on their device. Building mass surveillance features into the device only makes that easier to accomplish than before.
But everybody agrees that outlawing the possession of CSAM in general is the right thing to do. This was the case even in a time when the Internet was still in its infancy.
On the contrary, with passive detection you don't have to risk tipping the person off to the police yourself, so it becomes even easier to frame people.
By crossing your fingers and hoping that the Apple reviewers mistake the grey blobs for porn, the NCMEC reviewers mistake the grey blobs for porn, and the police, DA, and judges all mistake the grey blobs for porn, and that the person being targeted doesn't notice the mysteriously appearing grey blobs on their phone in time for that entire process to happen?
And why would anybody be more interested in framing somebody with an image that looks so similar to child porn that it convinces dozens of professional child-porn investigators and yet technically isn't, instead of just sending the actual child porn?
They neeed to already have the original image to make the hash collision. In your scenario, what does the attacker gain from sending a visually-indistinguishable collider instead of the original?
Yes, it would, because they need to have the original CSAM image in order to forge the hash collision. This method doesn't work unless you have the original photo to look at, meaning that your device will be raising alerts just as much as the targets, except you will be the only one incriminated since the target doesn't actually have any CSAM, only the hash-collided fake that you've made.
You don't need to, you only need the hash itself to create a collision. Which means that you only need a malicious actor to steal and/or sell the hashes.
And how then are they going to make an image that is visually indistinguishable from the original?
Remember: the hash collision is only the first step in the process. You also need to convince the Apple reviewers that the image is pornographic, and convince the NCMEC reviewers that the image is in fact the same as the one that it met a hash match with (apparently by magic, since you're imagining that the original image that you have to make a visually-indistinguishable match to isn't available).
I ask again: how is this process any more potent than any regular old fashioned pornography framing?