Aardvark Daily

Please visit the sponsor!

Illusion or effective CSAM blocking?

Anything that reduces the amount of CSAM on the internet is a good thing.

AI systems are now apparently doing a good, but not perfect, job of helping identify such material and report those who traffick in it.

Unfortunately, sometimes these anti-CSAM systems are a little over-enthusiastic, as witnessed by the case where someone had their engire Google account deleted by AI after simply sending some photographs of their ill child to a doctor.

Apparently there have also been issues with people creating harmless AI images with Google's Nano Banana, only to have them flagged as CSAM and the user's accounts deleted.

All this aside, I was rather surprised to see an NZ ISP jumping on the anti-CSAM bandwagon in a way that I can't quite fathom.

The ISP is Spark and it's claiming that it will be implementing technology to block CSAM material.

It's not so much focusing on actual CSAM material involving minors but on AI-generated CSAM images.

According to that RNZ story, "The company already stops people accessing photorealistic material made with AI".

Eh?

If it is planning to block AI-generated CSAM images then what is the "photorealistic material" it currently blocks and why?

Oh... it seems that this new initiative is designed to block "non-photorealistic AI and computer-generated child sexual abuse material, including cartoon-style imagery".

Actually this whole release/story is kind of confusing and deceptive IMHO.

For a start, Spark isn't actually blocking AI-generated CSAM material itself, it's simply adding some URLs to a block-list. Those URLs point to sites that have been identified as hosting such offending material so there's really nothing new here, they've been doing that for a long time.

The news story gives the impression that there will be smart AI stuff going on that allows Spark's systems to identify and block individual images -- but those of us who know how the internet works immedialy realise that this is impossible.

Why is it impossible?

Well almost all web and other internet traffic these days is encrypted using TLS. This encryption makes it impossible for third parties (such as an ISP) to intercept and monitor the data flowing through their systems.

It's also worth noting that I guess anyone who was hell-bent on accessing this CSAM material would simply us a VPN or find a suitably vulnerable proxy-server on the net and bypass the relatively basic roadblocks implemented by Spark and other ISPs.

On the plus-side, I guess this will stop the casual CSAM-user and maybe prevent the unwitting from stumbling across such horrific material by accident but I'm not sure that pronouncements such as the one from Spark don't actually create something of a false sense of security.

Carpe Diem folks!

Please visit the sponsor!

Here is a PERMANENT link to this column

Rank This Aardvark Page

Change Font