> CSAM can be detected through hashes or a machine-learning image classifier (with some false positives), whereas

Everything can be detected "with some false positives". If you're happy with "with some false positives", why would you need any context?