Why CC signals?

[CC-signals]

Thanks to everyone who shared feedback so far. We’re listening and this is just the beginning of the conversation. This update responds to early questions about CC signals. A more comprehensive summary and next steps will come after the current consultation period has ended.

Thanks to everyone who attended our CC signals project kickoff last week. We’re receiving plenty of feedback, and we appreciate the insights. We are listening to all of it and hope that you continue to engage with us as we seek to make this framework fit for purpose.

Some of the input focuses on the specifics of the CC signals proposal, offering constructive questions and suggesting ideas for improving CC signals in practice. The most salient type of feedback, however, is touching on something far deeper than the CC signals themselves – the fact that so much about AI seems to be happening to us all, rather than with or for us all, and that the expectations of creators and communities are at risk of being overshadowed by powerful interests.

This sentiment is not a surprise to us. We feel it, too. In fact, it is why we are doing this project. CC’s goal has always been to grow and sustain the thriving commons of knowledge and culture. We want people to be able to share with and learn from each other, without being or feeling exploited. CC signals is an extension of that mission in this evolving AI landscape.

We believe that the current practices of AI companies pose a threat to the future of the commons. Many creators and knowledge communities are feeling betrayed by how AI is being developed and deployed. The result is that people are understandably turning to enclosure. Eventually, we fear that people will no longer want to share publicly at all.

CC signals are a first step to reduce this damage by giving more agency to those who create and hold content. Unlike the CC licenses, they are explicitly designed to signal expectations even where copyright law is silent or unclear, when it does not apply, and where it varies by jurisdiction. We have listened to creators who want to share their work but also have concerns about exploitation. CC signals provide a way for creators to express those nuances. The CC signals build on top of developing standards for expressing AI usage preferences (e.g., via robots.txt). Creators who want to fully opt out of machine reuse do not need to use a CC signal. CC signals are for those who want to keep sharing, but with some terms attached.

The challenge we’re all facing in this age of AI is how to protect the integrity and vitality of the commons. The listening we’ve been doing so far, across creator communities and open knowledge networks, has led us here, to CC signals. Our shared commitment is to protect the commons so that it remains a space for human creativity, collaboration, and innovation, and to make clear our expectation that those who draw from it give something in return.

Our goal is to advocate for reciprocity while upholding our values that knowledge and creativity should not be treated as commodities.

Our goal is to find a path between a free-for-all and an internet of paywalls.

Copyright will not get us there. Nor should it. And we don’t think the boundaries of copyright tell us everything we need to know about navigating this moment. Just this week, Open Future released a report that calls for going beyond copyright in this debate, on the path to a healthy knowledge commons.

This is the beginning of the conversation, not the end. We are listening. From what we have heard, CC signals, or something like it, is the best practical mechanism to avoid the dual traps of total exploitation or total enclosure, both of which damage the commons. We have shared our current progress because we want to learn how to design it to meet your needs. We invite you to continue sharing feedback so we can shape CC signals together in a way that works for diverse communities.

In the months ahead, we’ll be providing more detail about how CC signals are developing, including key themes we are hearing, along with the questions we are exploring and our next steps.

[ghost]

But why so few "nuanced" options and none of the ones creators are asking for? How is that agency? That comes across as a "Any color the customer wants, as long as it’s black." statement.

[rebeccakross]

This is a great question and one that we've struggled with ourselves. With the kickoff, we had a list of "big questions" we were hoping for feedback on and one of them states:

"CC signals prioritize standardization and machine readability to increase their utility, but this comes with costs. How do you think we should consider the tradeoff between context-specific considerations and the goal of collective action that spurs behavioral change by reusers operating at a massive scale?"

The short answer to your question is that we opted for fewer signal options to prioritize machine readability, interoperability, and to streamline the creation of datasets for data collection holders. This is obviously an imperfect solution and one we will continue to explore as the initiative evolves.

[tuxwise]

"The most salient type of feedback", as you put it, was not what the above gaslighting and reframing wants to suggest, but the simple request to stop this "project", and to not sink your reputation that was built by many believers and promoters (including me), in the course of many years.

You're making it clear that you intend to proceed, no matter what.

So be it, I'm done with "CC" licensing. Bye.

[jcolag]

I still say that you're looking at this from the wrong end.

The problem is that we don't have a convenient enough way to offer up our work to a billion-dollar company that has no interest in paying for labor while expressing our wish that they would maybe help out. The problem is the billion-dollar companies that have no interest in paying for labor.

Who is this for? I can maybe see Mozilla's Common Voice using it, but literally nobody else, because who both trusts that abusive corporations will care what we want, this time, and doesn't mind massive corporations profiting from their work?

[cadadr]

Thanks to everyone who shared feedback so far.

I did not share any feedback. I criticised this ill-advised nonsense of a "project".

Some of the input focuses on the specifics of the CC signals proposal, offering constructive questions and suggesting ideas for improving CC signals in practice. The most salient type of feedback, however, is touching on something far deeper than the CC signals themselves – the fact that so much about AI seems to be happening to us all, rather than with or for us all, and that the expectations of creators and communities are at risk of being overshadowed by powerful interests.

These are plain lies, gaslighting indeed as @tuxwise puts it. This was not what was said. People said "AI is a vague marketing buzzword" and showed no desire for "it" to "happen" "with or for us all". Creators, and seemingly whatever "communities" might be, don't have "interests" that are "overshadowed"; they seem to desire that this nonsense just get off their back and leave their work alone. Your wording in general redressed all the backlash you got and the backlash so-called "AI" got as disappointment with some management decisions and product feedback

Honestly, this kind of engagement is so off putting in and of itself that I don't feel like ever using or recommending your licences again merely on that basis, regardless of what you might do with this so-called "project" eventually.

CC signals provide a way for creators to express those nuances. The CC signals build on top of developing standards for expressing AI usage preferences (e.g., via robots.txt). Creators who want to fully opt out of machine reuse do not need to use a CC signal. CC signals are for those who want to keep sharing, but with some terms attached.

LLM companies are violating any measures of this sort, up to and including attempts to block their access. This is pie in the sky thinking at best, but the realistic outcome is the name of CC is sacrificed as PR fodder for them. And along with allowing them to deploy the name and reputation of CC, all CC users' past usage and endorsement of both the licences and the institution are rendered at their feet along with it. All you're doing amounts to awarding them a safe way to say "you're withholding consent wrong, so your non-consent is invalid".

Our goal is to advocate for reciprocity while upholding our values that knowledge and creativity should not be treated as commodities.

Reciprocity between authors and artists and global trillion dollar companies?

From what we have heard, CC signals, or something like it, is the best practical mechanism to avoid the dual traps of total exploitation or total enclosure, both of which damage the commons.

Heard from whom?

[greg-kennedy]

The CC signals build on top of developing standards for expressing AI usage preferences (e.g., via robots.txt).

I keep seeing robots.txt thrown around as proof that extralegal standards can work, but it's a very poor comparison.

The reason robots.txt worked is that it was mutually beneficial: webmasters wanted bots to stop indexing certain things, and search indexers wanted to avoid polluting their datasets with garbage results. The balance led to better SEO and helped search engines better funnel results to websites. Win/win. So, yes, it was a "social" solution, but also reliant on the "technical" teeth to work. (And indexers that don't care about garbage but are looking for secret goodies - open directories, password files, admin panels - don't read it at all, or worse, use it as a hint to find the good stuff!)

If you're trying to do something similar with CC-Signals, I have to ask: what's the "stick" to accompany the carrot here? The whole thing is very vague on that front - some indicators of "well AI scraping that doesn't respect this is uhh going to suffer quality problems down the line!" but no concrete explanations. So AI scrapers won't care at all. Meanwhile, webmasters have no real incentive to adopt this either: Chatbots don't link back, they don't attribute, they don't drive traffic - in fact they go out of their way to obscure sources. If the implied default is "don't scrape my content" there's no reason for a webmaster to want anything different.

I believe this is because you don't have any disincentives to offer. You have no stick. Which is why this seems ultimately a doomed operation. I think that robots.txt arose "organically", like sitemaps.xml did and other ad-hoc standards to guide the web. This attempt to force something to happen from the outside is burning CC's reputation for no benefit. As I posted elsewhere, this is much more likely to go the way of Do Not Track. Are you sure you want to stake your userbase trust on such a dead end project?

[CC-signals]

We speak more to how we might incentivize compliance in this FAQ response.

First, there are contexts in which CC signals are likely to have legal effect. For example, the European Union’s legal regime on copyright and AI, particularly the Directive on Copyright in the Digital Single Market and AI Act foresee the possibility for rightsholders to opt out of the text-and-data mining (TDM) exception. In those cases, rightsholders reserve the right to conduct TDM and model developers need permission to use copyright works for AI training (see Article 4(3) of the EU Directive on Copyright in the Digital Single Market (EU 2019/790)). CC signals may serve as an express reservation, i.e. an opt out mechanism in this context, however there is still more to be determined.

Second, we have been exploring other measures to incentivize compliance, such as a petition to AI developers; an accreditation system where AI developers who follow CC signals get a special designation; and more.

[ghost]

The signal we want is simple: "no, sod off or get sued."

That simple signal doesn't appear in any of the manure you've shoveled out to us so far, CC.

Figure out how to make it appear and we might come back. As is, you're cooked. Your community is vanishing because you have refused to listen to it and serve it.

You're done, CC. Finished. Over. Kaput. Goodbye, from a two-decades supporter.

[TimidRobot]

We added an FAQ entry about opting out:

• Can I opt out of AI training altogether? #64

  Yes, there will be tools available to express a preference of no AI training. If a creator wants to opt out of AI training altogether, they should: 1) use robots.txt with IETF AI Preferences opt-out when it becomes available; 2) use a technology solution to prevent access; or 3) remove your works from the internet.

  CC signals doesn't include its own opt-out preference because the foundation of CC signals, the IETF AI Preferences, already does. Like CC signals, the AI Preferences are still a draft. For more information see the IETF AI Preferences (aipref) working group. There are also technologies designed to prevent AI from accessing content. See this comment for examples.

Also, based on similar feedback, we added opt-out to our README:

• Add opt-out information #58