upvote
Every mainstream vendor with a US presence does this on their ZDR/MAM AI endpoints. This also includes AWS and Azure. It's explained directly in the AWS ZDR docs, and it falls under the "synchronous" scanning pipeline in Azure with full MAM.

At least with AWS they're clear that this is the only type of proactive scanning they do on their ZDR AI endpoints. With Azure it's actually unclear exactly how much more falls into the synchronous scanning pipeline, though I may be missing some docs.

The way this was explained to me is although there is no legal requirement to do this, vendors are concerned that AI processing constitutes "actual knowledge" of the content unlike a lot of traditional cloud-based workloads, which would expose them to legal liability. Seems like a weird argument in a ZDR context, but anyway, there's also a public optics element too I'm sure.

reply
I'm personally hopeful that as an industry we can move towards E2EE and attested secure enclave pipelines to remove the "actual knowledge" argument. NEAR AI and Tinfoil are the only publicly available providers in this area but I would be excited to hear about more advancements in this area because I personally think that these NCMEC integrated "ZDR" systems are major privacy violations and deceptive marketing.
reply
Azure does offer nVidia's secure enclave architecture via "Azure Confidential GPU", but it's limited to a single GPU and a little less than 90GB after overhead, so basically maxes out at Mistral-Large-2 with 4 bit quantization.

This whole thing kind of puts the industry in a bind. On one hand, the big labs do have an incentive to get this kind of thing rolled out so as to avoid open weight on-prem AI gaining traction, and to remove remaining roadblocks to corporate adoption in regulated industries. But rolling it out risks attracting even stronger regulatory responses for them as well (that's even setting aside likely national security/law enforcement pushback), as well as likely being strongly opposed by the AI safety people. There's no one way to square the circle that will please everyone. My guess is the labs will take a third route: continue to lobby against open weight models, thus killing the possibility of on-prem, and then drag their heels on E2EE/secure enclave models except perhaps for military customers.

reply
I'm hopeful that Tinfoil's enclave router, which does decrypt the payload but does so in a verifiably secure manner, can be used to scale to multi-gpu models (for example, Tinfoil does offer Kimi K3.1 Pro).

Because of economies of scale with batch inference as seen with DS4.1 Flash, this is the only avenue for true privacy from frontier models at scale/cost.

In terms of policy, I agree it's looking dire. There's unfortunately just more money/lobbying/mindshare on the 'safety at all costs' camp but I'm still advocating and educating where I can.

reply
We will however need hardware elements that are actually secure. The current generation is trivial for the operator to compromise. https://tee.fail/
reply
deleted
reply
How are these companies able to legally train these models? Are the employees citizens of countries that this material is illegal to download or possess?
reply
Most CSAM statutes have a carve-out exemption for things like this.
reply
But they still have to obtain the images, look through the images and do tests with the images? Sounds like a totally fucked up thing to do. Reminds me of a fellow student that now works for a IT forensic company that does forensics for the police. He said that 98% of police requests involve checking storage of suspects for CP. He also has to look at the images/videos to write a report. He changed, no joke.
reply
i'm pretty sure certain organizations are given authorization
reply
there never was a point in using rent-seeking middlemen anyway
reply
Yeah, that's lovely. It's either ZDR or it's not, for Christ's sake.

"ZDR, except for _______" is not ZDR.

reply
While I don't support such measures zero data retention does not mean zero data inspection so I don't think there's anything dishonest going on. Also it's generally to be assumed that any records believed to be linked to criminal activity are retained regardless of industry so again not an issue as far as the advertising goes.
reply
I suppose you're right. It's hard to imagine them granting users unlimited capabilities.
reply