>The license goes beyond redistribution. You are granting a sublicensable and transferable right to your content, giving the platform the legal authority to sell or license it (or to not license it) to AI scrapers and other entities. The platform's right to block said scrapers comes from posession rights.

They don't need to sublicense it because the license was already granted by you. Stackoverflow comments are licensed under creative commons, which means you don't need to seek a license from stackoverflow to use it. It's same if you found some random MIT licensed repo on github. It's not github granting you a sublicense, it's coming from the original author.

>You still technically own the copyright, but the museum owns the building. They can lock the door, charge admissino, kick out anyone they want, prevent anyone they want from seeing it, etc.

And Anthropic can't decide who gets to use their service, and for what purpose?

▲

thewebguyd an hour ago | parent [-]

Anthropic can decide who gets to use their service. They have complete control over their services and service.

It still breaks down once the output has left the system though. Anthropic cannot tell you what you can and cannot do with the LLM's output, they do not own that, its public domain. Anthropic can pursue breach of contract, maybe, but they can't do anything regarding your use of the model's output. If China can't access Claude directly, they can just pay some other user in the states to run some prompts and paste the output on a public website, and then use that output and there is nothing Anthropic can do about it.

Fair point on StackOverflow, but they are the exception rather than the norm. Most social media doesn't license the content under creative commons.

▲

gruez 31 minutes ago | parent [-]

>Anthropic cannot tell you what you can and cannot do with the LLM's output, they do not own that, its public domain.

And are they actually doing this? For instance, if you read their press releases about distillation attacks[1], they're not asserting copyright over the outputs, only alleging "fraudulent accounts". So far as I can tell they're not even engaging in legal action.

[1]https://www.anthropic.com/news/detecting-and-preventing-dist...

	▲	thewebguyd 7 minutes ago \| parent [-]
		They aren't taking legal action yet, no, because they have no legal ground to stand on. But they are pushing lawmakers to do something [1] They are also constantly using the word "illicit" and theft in their communications, and in their lobbying, when there nothing illicit about using model output to train another model. They are trying to create an aura of criminality where none exists. > But distillation can also be used for illicit purposes: competitors can use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently. They do have leverage over fraudulent accounts, yes, but the resulting distillation from those is out of their control under the current legal framework. There's nothing they can do about it, for now. [1] https://arstechnica.com/tech-policy/2026/06/anthropic-claims...