[JayGuerette]

Apple isn't creating neural hashes for CSAM detection, as they'd have to be in possession of source material to create them, so they're getting them from someone else. Since it's indistinguishable in it's hash form, when the supplier becomes interested in looking for something else, nobody will ever know.

[echelon]

Do not let Apple off the hook. This must be removed.

This functionality will be used in other global jurisdictions to clamp down on freedom. In a world where we cede more control and increasingly subjugate ourselves, it's only a matter of time before it's used against us too.

Say no to monitoring.

[brookst]

So the thing the article concludes isn’t being done, must be removed?

[echelon]

Maybe that's the case. We should be vigilant and treat the concern with utmost seriousness.

Once upon a time, Apple announced they would do this. We can't ever let them.

[nostromo]

The article concludes that there isn't evidence it's being done -- which is different.

Apple should not be trusted without evidence. We should be able to verify what our devices are sending to Apple, Google and others.

[onethought]

But they aren’t monitoring. And a hash doesn’t help you monitor unless you’ve got a hash of another image like CSAM. And you can turn this off if you want.

Honestly I know everyone thinks that everything is a slippery slope to death camps. But sometimes it’s just a cool helpful feature.

[brookst]

The suppliers are well documented and it takes two suppliers agreeing on the same neural hash.

So, when the US center for missing and exploited children decides to collide with the Japanese equivalent to detect IDK what, yea, you wouldn’t know. Assuming those agencies don’t operate with transparency.

[rurp]

Requiring two suppliers to agree is simply the current policy. I think the GPs concern is that Apple's policies can change without warning or notice. That seems like a pretty valid concern to me, which Apple has zero interesting in mollifying.

[yreg]

>so they're getting them from someone else

Is there any evidence they even do neural hash CSAM detection?