delusional
5 hours ago
The court was asked to annul the previous judgment since the defendant did not think that the pictures in question were CSAM, since the children depicted were fictional. The court found the pictures similar to photography and therefore rejected the appeal.
In my mind, this is pretty substantial technology news, at least in Denmark, since it rejects the notion that AI generated pictures are substantially different from other pictures, and therefore has to abide by the same laws.
I will be interested in seeing if anybody charges the big AI providers for producing these photos.
andsoitis
5 hours ago
A fictional, generated image seems quite different to me than a photo of a real thing that exists.
So I think one has to find other logic to contain this crap.
delusional
4 hours ago
I might to you. They are not different as they relate to CSAM as a matter of Danish law.
andsoitis
4 hours ago
I recognize that. I think my point is one about internal consistency of a legal framework.
For instance, what if we're NOT talking about CSAM, are generated images ALSO considered equivalent to photos of real things in other domains? Or is CSAM a special carve out and based on what principles? Legal frameworks are stronger when they limit arbitrary rules.
(to be clear, there might be a good logic argument, e.g. it isn't about protecting children, it is about protecting adults - so it doesn't matter how the image came into being).
altairprime
an hour ago
The generalization proposed is invalid. Most other photographic domains do not rise to the level of censure levied at CSAM, for social and cultural reasons. Generalizing a CSAM-context outcome to non-CSAM contexts would be inappropriate, as there is no causal relationship between outcomes on either side of that chasm. One must instead judge if actions taken to prevent CSAM will present undue harms in non-CSAM contexts. This is the ‘carve-out’ you consider the existence of, and its boundaries are as much defined by each culture as they are by each legal system.
So, then: What specific and immediate — not potential, but actual — harms does this CSAM-context decision impose on non-CSAM contexts?
At a glance, I don’t see any such harms imposed. One could perhaps argue that AI must take more care now, and that this is a harm; plausible enough, but is that an undue harm? Not likely: the cultural and legal hostility to crossing from non-CSAM contexts into CSAM-context has no reason to yield to the AI operations difficulties imposed by restricting generation of CSAM-context materials.