Suppose Apple flags some image in iCloud using a hash that was probably sent from the device. Do they currently have the capability to decrypt the image to check if the content is actually illegal before reporting it to the authorities without needing access to the physical device?
(Not trying to make a counter-argument to your comment. I genuinely don't get this part).
Hmm not sure what you mean. The hashes are shipped with iOS, and then (if parental controls are enabled) compared with the hashes computed against the unencrypted data on-device, and displays a warning to the user and their parents.
I'm not sure how the image is displayed on the parents' device - it could be sent from the child's device, or the image used to compute the hash that the child's image matched could be displayed.
(Not trying to make a counter-argument to your comment. I genuinely don't get this part).