I hope you do not take my ideas as a personal assault (they are not), I’m reacting to a movement that is trowing copyright into the trash bin. The legal framework is broken so we can only rely on keeping the ethical norms up ourselves (if we want to keep something of copyright alive).
My clear answer to your question would be “no AI” also apply to rendered images (mp4, png, etc.). The terms say, quote<<“Licenses marked with "no AI" tag do not permit “Product use” for machine learning or training of neural network models, including generative AI models>> end quote.
I believe “product use” means product use, the use can be whatever, e.g. decompose the data (product) to other format, render/rasterize, voxelize, etc. Non of these data resulting from the use can enter a training dataset or enter generative AI models.
(I would argue that is when the work is still recognizable when incorporated in a scene and even more certainly when it is standing on its own)
I also believe this would have been common sense a few years ago, but some companies decided to start an infinite public discourse about fair versus unfair use and made international copyright law a moving target. They also keep adding all sorts of reasons to the discourse so it can go on indefinitely (how they like it). So now when someone says, pleas do not use my work this or that way, it seems not that common sense anymore to honor it. That is why the question on top of this post sounded like nails scratching on a chalkboard to me and it probably colors my comments here so don’t take it personal.
I hope my comments (buzzing with some frustration) somehow wake readers and prompt them on reflecting on how this can possibly get to a positive outcome?
In few months from now (maybe year tops) the only thing one is going to need is images and then this CGTrader repository is basically free to download, just copy paste the presentation images into an AI and there you have the 3D model, the only thing left holding this back is (fading) common sense.