4

There is a lot of buzz now around word AI, and many security aspects of AI technology and which is more importantly AI application remain unclear for me. This is just a small subset of such questions that came to my mind:

  • How safely create LLM models not infringing anybody's copyright?
  • How to protect my potential digital assets from AI scrapers? I.e. which platform to choose
  • How to protect my already existing assets created on different platforms:
    • code: on Github-like platforms
    • images: on IG and alike image storages and banks
    • videos: on YT and alike videohostings
    • etc.
  • How to create digital watermarks to protect my assets from AI scraping?
  • How to save and keep stored safely the datasets for LLM models?
  • Technologies and framework to protect against AI scraping? Something like Cloudflare is creating now

Which of that is off-topic here and why? Thanks in advance.

1 Answer 1

3

AI security questions are certainly on-topic, however some of your examples are not security questions or a good fit for our site.

  • copyright infringement is not a security concern
  • recommendations for a platform or protection tech is a product recommendation question and off-topic, so is "which platform doesn't process my assets with AI?"
  • how to protect existing assets is far too broad the way you've approached it, because the problem isn't where the assets are stored, but how they are processed, so those questions would need refinement so that they weren't opinion-based or unfocused

As a complete tangent, watermarks do not protect assets from scraping, they are a detection control, and I don't think that approach is a viable subject, but, as I say, that's a tangent.

As with every new threat, always be careful of the question pattern, "I'm afraid of X, would [arbitrarily chosen Y] protect me?" as they often end up in a situation where the entire premise needs to be challenged at the foundational level, and therefore unanswerable as asked.

4
  • 1
    copyright infringement is not a security concern agree, but the content became copyrighted only when it was registered with a trademark or patented, which is not the case for many digital creators in the Internet. The content created on different platforms is often governed by their blurry T&C and EULAs, so it is a natural intention to protect it by own means, this is what my questions are about. I found that DRM questions are ontopic here, and above are closely related to DRM, isn't it?
    – Suncatcher
    Commented Oct 6 at 0:53
  • Also I found multiple torrent and p2p questions, which is also somehow related to file sharing and prohibition of such sharing
    – Suncatcher
    Commented Oct 6 at 0:53
  • 1
    Read my answer to that linked DRM question. Many questions that might touch on a certain topic domain don't end up being about that domain, and that's what I'm anticipating that you will end up with if you want to ask questions about copyright infringement. Don't take a topic that you find is on topic and shoehorn in a question you want to ask because you think you can justify the question based on the fact that other questions within a topic domain are answered here. The first test: is the question about Security? The second test: what's the topic domain? Don't do it the other way around.
    – schroeder Mod
    Commented Oct 6 at 8:07
  • 1
    Copyright infringement is inherently a legal topic. How to avoid infringement when designing an AI model is a programming question. How to avoid infringement as a user of an AI model is a question for the model and its designers.
    – schroeder Mod
    Commented Oct 6 at 8:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .