This would hypothetically be a possible way, but there’s two issues with it:

1. False positives could never be verified, and as such, if you are the victim of one, it’s your word against theirs. And they usually control whatever you’re trying to access, so they will take the benefit of their doubt and block you anyways.
2. These companies have very little incentive to go this way unless it saves them money, since this requires more expensive engineers to build and maintain. Afaik it’s always been cheaper for them to just circumvent what little regulations there are and do things the easy way.

EDIT: But yeah as long as it’s not used for any kind of security measure it should be fine.

While I agree that the AI they will implement will likely not be very effective, it doesn’t have to be to cause massive human suffering. Eg. Google incorrectly marking exposed photos of your kid for your doctor as CSAM. There’s also no guarantee that once these companies finally wake the fuck up (If they’re not already completely aware what they’re doing is messed up) that they will close these holes they’re punching, and that could mean they could replace AI with a mass surveillance tool at any point without you knowing. Nobody should be a fan of this.