Oct 8, 2026, 5:48 PM@cwebber @davidrevoy Isn't this what CC BY-ND is for? 🤔
- Oct 8, 2026, 5:48 PM
Replies
Oct 8, 2026, 6:03 PM@dajb @cwebber @davidrevoy I'd think it would be easier to start with CC BY-ND and then add an exception to allow for derivatives if they were not AI.
However, I am not a lawyer. Also: possession is nine-tenths of the law, and the AI companies have already trained on the data.
Laws, contracts and licenses are only useful if all parties actually agree to the terms. The written text of the licenses is really only to avoid confusion about the agreement. But the AI companies don't show good faith.
- Oct 8, 2026, 7:03 PM
@poleguy @cwebber @davidrevoy Well, also, although you can relicense, anything you've originally released just under CC BY is fair game.
Cassandrichdalias@hachyderm.ioOct 8, 2026, 10:19 PM@dajb @poleguy @cwebber @davidrevoy CC-BY does not allow AI training use because such use does not preserve attribution. It's great to be explicit about that, but there is no "the old license allowed it" loophole here.
- Oct 9, 2026, 7:05 AM
@dalias @poleguy @cwebber @davidrevoy no
If only it were that simple:
"CC licenses apply only when copyright permission is required. If exceptions or limitations apply, then the CC license terms don’t apply."
https://creativecommons.org/using-cc-licensed-works-for-ai-training-2/
Oct 9, 2026, 12:19 PM@dajb @dalias @poleguy @davidrevoy I worked at CC, am familiar with what the licenses say. And attribution is still a requirement: if it's going to be a requirement to respect proprietary licenses going into datasets, it will also be a requirement to preserve attribution. The thing I am pushing on here is that whatever copyright outcome we get in terms of LLM ingestion and generation of content needs to be *equitable*, and on that front, attribution remains a requirement. You can't just hoover up the commons ignoring the few requirements it does have because it's the commons.
- Oct 9, 2026, 12:37 PM
@cwebber @dalias @poleguy @davidrevoy
IANAL and I haven't had the privilege of working at CC (merely being a Fellow) but I have seen a lot of hand-wringing on mailing lists and in blog posts.
Ultimately, however we might wish it otherwise, the consensus seems to be that if governments consider training data not to subject to copyright considerations, then neither do derivative licenses apply.
Again, I'm a mere observer. I may be wrong.
- Oct 9, 2026, 4:39 PM
@dajb @cwebber @dalias @davidrevoy
I agree, that's what's happening!
Generally speaking: We are the people. The government is us.
The government is not "they." It's 'we.' If we don't accept this madness, we should not blame it on the government, or 'them' and roll over in a ditch, but we should take back control from the tech lobbyists and make the government match the will of the people.
How to do it... not so simple.
RobotistryRobotistry@fediscience.orgOct 8, 2026, 7:09 PM@dajb @cwebber @davidrevoy Please correct me if I'm wrong, but I think the problem is wanting a CC-BY-ND-ForAI combined with a CC-BY-ForNotAI so the models aren't allowed to change or use the image while the humans are?
- Oct 8, 2026, 7:51 PM
@Robotistry @cwebber @davidrevoy It's wishful thinking, sadly.