Diffusion + privacy

If I only use diffusion to get quick realistic rendering using rendering presets available (like interior realistic) will generated rendering will become public domain? If yes, what does that mean exactly?

I think this is not a question about privacy but of ownership…

definitely ownership too. For example, I have a scene that has a logo and not yet public imagery and I use diffusion to generate a realistic rendering, will this rendering somehow would surface on the internet if someone’s searches.

I really don’t know the answer to that.
Maybe someone from SketchUp knows.

a lawyer.
no SU employee will (or should) answer that, because that would make them and the company liable.

if you have questions about copyright, both in your country and international, you should consult a lawyer.

but being CC0 doesn’t mean that it can’t be private.
Any image made available under the CC0 license has no copyright, is not “owned” by anyone, and is freely available to copy, modify, or distribute (even for commercial purposes), with no permissions needed.
Still, if you don’t make the images widely available to to world, who else will do anything with them ?
And let’s be honest here, both the NFT rush and the AI “oh well we trained them on billions of images found online, sorry” trends shows that in the end, regardless of the copyright, people will do whatever they want with online content.

the only difference is that here, it’s legal.



So, is it safe to say that if I don’t share generated image on line it will not somehow appear on line?
Meaning if I create a rendering with help of diffusion, save it and share internally within my company it will not somehow appear on line?

I seems to me that this question implies that the results of SketchUp Diffusions might be made public by SketchUp?
I think it’s a valid question…

yes, that is my question. Will Sketchup make it public images that are created by using diffusion? when i render without diffusion, renderings are not public.


What is the copyright for using Stable Diffusion-generated images?

The area of AI-generated images and copyright is complex and will vary from jurisdiction to jurisdiction.

What is the copyright on images created through Stable Diffusion Online?

Images created through Stable Diffusion Online are fully open source, explicitly falling under the CC0 1.0 Universal Public Domain Dedication

they technically could. So could I. or anyone who gets access to them. But as they state on the link Kyle gave, the copyright thing is complex and varies from juridiction to juridiction.

I’m not saying that as a “please stop asking questions” thing. I’m honestly saying “for a more complete answer, contact a local copyright lawyer of sorts”
any other answer, yours, mine, will be pure conjecture. I mean, I don’t even know where Oplavins is from, how could we give educated legal answers ? :smiley:

the main thing I’m seeing here is that that way, they can feed the images back to the AI training model. without having to ask for permission. Because looking at the big midjourney class action building up, the time where AI devs would just vacuum the internet for images is coming to a pause.

because you use softwares that grant you the copyright from your own images.
The tech solution behing SU diffusion was made available as CC0, therefore all applications like SU Diffusion have to abide by the rules.

the only unclear thing @KyleB , maybe you can bring some light, is a wording thing :
In your link there are mentions of “Stable Diffusion” and “Stable DIffusion Online”.

are there different rules depending on where you make the image ? in SU GO / iPad / Pro ??

1 Like

@ateliernab That I do not know. I am not on the dev team.
I would assume they are the same thing since you need to have an internet connection to use Stable Diffusion. Google tells me “Stable Diffusion Online” is the product name - https://stablediffusionweb.com/

1 Like

Yeah, thinking of it, even in Pro it’ll work on the servers (the “robots” are at work), so you’re probably right, it’s all the same. And regardless, it’s based on the same tech in either case.

this is the problem with tech and product names that use a generic word like diffusion :smiley:

I would not use Diffusion for a realistic rendering… Not yet at least…

We are NOT using our users generated images in the AI training model.


NO… As you create your images using Diffusion, the images are not published to the public. These are delivered only to you.


noted :slight_smile:
today, you won’t. but the CC0 licence of the diffusion tech technically allows you to, since they are public domain images :wink:

(and on an ethical note, using the images made with your AI tool to train the AI tool is way more morally just than using images vacuumed online as all of them are doing. it wouldn’t be shocking.)


Thank you, Chris!

As to the privacy part of it, using anything that uploads anything through a web connection with truly critical projects would be a strict no-no.


Hi all, and thank you for everyone’s contributions to this important topic.

I want to clarify that:

  • SketchUp Diffusion is NOT “Stable Diffusion Online”, which is a completely different tool and is unrelated to our services.
  • SketchUp Diffusion does not impose a CC0 license on the output.
  • A contributor to this confusion may have been the mistaken reference in our documentation to the Stable Diffusion Online tool, which we have now removed. Apologies for the confusion and thanks for helping us discover and fix that.
  • SketchUp makes no claim as to the ownership of the output from SketchUp Diffusion. Users own whatever rights are available in the output. That said, and as has been noted here, the area of AI-generated images and copyright is complex and will vary from jurisdiction to jurisdiction and we would encourage you to look into the copyrightability of AI-derived images in your jurisdiction.

For more information about the U.S. Copyright Office’s position on the copyrightability of AI-derived images, please see copyright.gov/ai.


But would such a thing even train an AI? Wouldn’t an AI be like a dog eating its own vomit if it trained on its own generated images? Meaning the quality would decrease with each iteration if it did that