Only if your daughter’s nudes end up in NCMEC’s database, which happens if (among other things) your daughter confirms to NCMEC that they were generated through an abusive scenario and she would like them classified as such.
But that's using "perceptual matching" too isn't it? They say it's configured for a low risk of false-matches, though they don't detail their testing procedure. Presumably they would know almost immediately of false matches, maybe if the odds are really 1/trillion they should commit to disabling the system if any false matches happen at all.
Also, is this all configured on the server-side? Suppose Apple has a pervy employee who vouches my daughter's account "looks suspicious," would he/she then be able to override or "manually review" all her pictures? Would I be notified of any such review?
It seems like none of your concerns are unaddressed. The following is all from the linked article:
1. Nobody new can see images sent over iMessage under the proposed changes.
2. A perverted employee cannot flag an account as suspicious; only a device can flag itself when a threshold number of photos matching the NCMEC database are found.
3. Accounts which meet the threshold and have the photo thumbnail verified as objectionable will be suspended with an opportunity to appeal.
No you didn't address my first concern. Which is, it's very easy for a company to write a blog post full of promises saying anything they want (e.g "1 in a trillion chance of false positives"), but it's all just words without some objective source giving oversight into the process.
Take for example point 2. According to what they've written here, yes that's how it works. But the point is, you have no flipping clue what's actually true, just what they typed. There could be debug modes, the false-positive numbers could be entirely incorrect, they never said what the "threshhold number" is, so that could be 1. They could have even meant everything that they said, but there could be a bug in the code.
Suppose there is a bug they discover, and the false-positives are exponentially higher than they quoted. Do you think they would publicly admit that? Also how big are these thumbnails and under what circumstances do they have access to them?
