Your Work Is Not Free Training Data
Reach is not ownership. Creators need proof, identity, and control over what happens to their work.
For a while, that deal worked well enough to feel fair. YouTube gave creators reach. Instagram gave photographers, designers, educators, dancers, musicians, and small brands a place to be seen. TikTok turned unknown people into public voices overnight. Substack, Patreon, Twitch, Spotify, and other platforms gave creators ways to publish, earn, and speak without waiting for a studio, label, publisher, broadcaster, or gatekeeper. Basically, each of these platforms opened a door that used to be guarded by someone else. And lot of real careers came from that change in the digital space.
However, to think that the old gatekeepers just disappeared would be…naïve, to say the least.
First of all, there was always a catch. Creators could build their audience, but the platform controlled the reach. They could upload the work, but the platform controlled the rules around visibility, monetization, discovery, analytics, and access. They could own the copyright on paper and still depend on a private company for almost everything that made the work commercially alive.
So, there always was a sort of dependency in the relationship creators — platforms. Now AI has made that dependency and the imbalance impossible to ignore.
Because now it’s not just about the fact a platform can suppress a post, demonetize a video, change the algorithm, or lock you out of your account anymore. Now, we have to put on the table the fact that human work can be absorbed into AI systems, turned into training value, and used to build products that compete with the same people who created the original material.
We already have examples of how ugly this can get.
In early 2026, folk musician Murphy Campbell found AI-generated versions of songs she had performed on YouTube uploaded to Spotify under her name, without her consent. According to The Verge, she suspected someone had taken her YouTube performances, created AI covers, and pushed them into streaming platforms as if they belonged there. Then it got worse. [S0]
A separate user, operating under the name “Murphy Rider,” uploaded private videos through distributor Vydia. Those uploads were then used to file ownership claims against several of Campbell’s own YouTube videos.
So the original artist found herself fighting claims against her own recordings.
The claims were later released, and Vydia said the user responsible had been banned. Vydia also said the copyright claims were separate from the AI-cover uploads. [S0]
And Campbell is not an isolated warning. Visual artists, publishers, record labels, actors, musicians, photographers, and writers are now pushing back in court and in public because the same question keeps returning: who gets to use human work as machine training material, under what permission, and who gets paid when that work creates value?
So, for creators, the warning is hard to miss: when platforms, automated claims, weak identity checks, and poor proof of origin decide who owns what, the person who made the work can still end up treated like the infringer.
Actor Joseph Gordon-Levitt has been one of the clearest public voices on this. In his recent work around AI and creator rights, he argues that many AI companies built powerful systems on human work without permission or compensation. At Axios’ AI+ Summit, he framed the economic problem directly: the humans who produced the data used to train these models hold “0% of the economic value,” while the tech companies can end up holding “100%.” [S1]
The way we often hear AI companies talking about “data” as if it appeared naturally, like weather, can be misleading. That data is definitely not random information. We’re talking about preferences and particularities, all those small decisions that are unique to a creator’s voice and make their work recognizable.
A creator’s archive is not just a pile of files. It’s years of struggle, research and refinement and creative choices that shaped the work and earned the trust of an audience.
So when companies say they trained on “data,” what creators hear is “you trained on our work, on our voice, on our audience and history.”
Gordon-Levitt helped launch the Creators Coalition on AI, alongside Daniel Kwan, Natasha Lyonne, Janet Yang, and other people from film, writing, acting, production, technology, and independent creative work. [S2]
Their starting point is that creators have been supplying the raw cultural material, while many of the companies building AI systems keep the control, the product, and most of the upside. That is why the coalition is pushing for consent, control, compensation, transparency, job protection, and safeguards against misuse.
This in no way means creators are against AI technology. Many creators already use it to test ideas, organize material, clean up production, speed up boring tasks, thus saving time for the actual meaningful work.
But this cannot become an excuse for extractive systems.
A creator can support AI and still reject the idea that their work should become free training fuel for a company that then profits from it, while treating the human who made the original work like a cost problem.
Coming back to our platforms issue, YouTube introduced a setting that lets creators and rights holders allow third-party companies to use their videos for AI training. The setting is off by default, and creators can choose which companies are allowed. YouTube also says it is not currently facilitating payments between third-party AI companies and creators or rights holders. [S3]
So, maybe at least some platforms are paying attention and understand that training permission is a big issue and creators want control over whether their work is used to train outside models.
It also shows that the platform still designs the permission system and defines the options.
Meta tried a different angle in Europe. In 2025, it announced plans to train AI models using public content shared by adults on Meta products, along with interactions people have with Meta AI. EU users can object, and Meta says private messages with friends and family are not used unless someone shares those messages with AI features. [S4]
That sounds more controlled than an open scrape of everything. However, that “public” term is still a big stretch.
A creator may post publicly because they want to reach an audience but that doesn’t automatically mean they’re OK with helping train a model.
Publishing and permission are not the same thing!
The legal system is now acknowledging the difference and trying to catch up with it. The U.S. Copyright Office released the third part of its AI report in 2025, focused on generative AI training. The report opens with the central questions: whether copyrighted works used in AI development require consent or compensation, and how that could be practically handled. It also notes the scale of the fight: more than 10,000 public comments, pending lawsuits, and global disagreement over fair use, licensing, and opt-outs. [S5]
In Europe, the AI Act has created obligations for general-purpose AI providers, including transparency and copyright-related requirements. The European Commission says the rules for general-purpose models include documentation, copyright compliance, and public summaries of training content. [S6]
Performers have pushed especially hard because AI can now imitate the actual person, not only reuse their work. SAG-AFTRA has secured AI protections across several contracts, including consent and disclosure requirements for digital replicas, voice protections, and compensation rules in specific agreements. [S7]
So, the pressure is coming from all sides now: creators, actors, musicians, writers, regulators, unions, and platforms themselves. And the demand is not complicated.
Ask before training on someone’s work, say what was used and pay people when their work creates value. Give creators a real way to say “no”.
Without that, the platform bargain breaks down completely.
For years, creators lived under the illusion that they owned their content. Well, technically speaking, they did.
But practical ownership is much weaker when almost everything around the work including the audience, payments, analytics, identity, archive, reach, and account access — belongs to a platform account they don’t actually control.
As a creator, you can own the work and still lose the audience or own the copyright and lose monetization. Furthermore, you still have to accept a new rule or policy update, or a new use of your public uploads for AI training.
That’s not ownership in any meaningful sense.
AI exposed this at another level because AI doesn’t only take from the finished post. It learns from the accumulated presence around the creator — their work and style, the audience response, the metadata, etc. It forces us to look at digital ownership in a more serious way than “it’s my account” or “my name is on the upload.”
For creators, ownership should include identity, proof of authorship, access to their audience, records of publication, permission around reuse, and a clearer path to compensation when their work creates value for someone else’s product.
At SourceLess, this topic is central to our work around owned digital identity.
We meet more and more creators at events, conferences, and inside our community who are asking a very real question: “What actually belongs to me online?”
The conversation may start from content ownership or from account dependency, AI and proof of authorship. Sometimes it starts directly from the fear that someone can take years of work and make it look like it belongs somewhere else.
However it starts, it’s usually around the same concern — creators are beginning to understand that the digital space they helped build is being rewritten around them.
Content moves faster now. It can be copied faster, reused faster, claimed faster, trained into models, and distributed again without the original creator being properly seen, asked, credited, or paid.
So digital ownership has to become practical. And people need to understand what they are giving away when they build everything through accounts they don’t own.
For a creator, ownership should not mean only having a username on a platform or a copyright statement buried somewhere inside an account dashboard. It should mean a stronger connection between the person, the work, the proof of creation, the history of publication, and the permission around future use.
That’s the direction we’re working toward.
STR Domains are designed as owned digital identities inside the SourceLess ecosystem: readable names recorded on-chain and connected to access, wallet, communication, creator profiles, and other tools.
Proof of Existence can help timestamp and log original work.
wNFT infrastructure adds a digital ownership layer that can support transfer, licensing, and future creator tools inside the ecosystem. [S8]
Of course, a timestamp will not solve every copyright dispute.
It will not force an AI company to pay creators tomorrow and certainly, it won’t replace law, licensing, contracts, unions, regulation, or public pressure.
But it gives creators something they badly need: a record that does not depend only on platform memory.
When work is logged, linked to an owned identity, and connected to a system built around proof and permission, the creator is no longer starting from the weakest position.
They can now point to a verifiable record attached to an identity they hold.
That changes the base from which creators publish, protect, license, and build.
For years, the internet trained creators to chase distribution first: reach, views, subscribers, algorithmic visibility.
And of course, distribution matters. Without it, your work disappears.
But if the AI fight makes one thing painfully clear is that distribution without ownership can become a trap.
The more creators upload, the more value they create for the platforms around them. If that value can then be absorbed into AI systems without permission or compensation, the creator becomes both the supplier and the training set for products they don’t control.
That is the part creators should not ignore.
AI can support human creativity. It can lower production barriers, speed up boring tasks, open new formats, and help independent people make things that used to require larger teams.
But if the economic structure is wrong, the technology will not feel liberating for long.
Human work has value before a machine learns from it.
A creator’s voice, archive, style, audience, identity, and digital history should not become free infrastructure for someone else’s product.
Creators built the internet people actually care about.
They shouldn’t have to rent their identity, lose the trail of their work, and then watch the next generation of platforms turn that work into private AI capital.
Resources / Sources
[S0] The Verge — “A folk musician became a target for AI fakes and a copyright troll”
https://www.theverge.com/entertainment/907111/murphy-campbell-folk-music-ai-copyright
[S1] Axios — “Joseph Gordon-Levitt: ‘Not a punk rock thing’ to use artists’ work to train AI for free”
https://www.axios.com/2025/06/04/joseph-gordon-levitt-ai
[S2] Creators Coalition on AI — Official website / coalition principles
https://www.creatorscoalitionai.com/
[S3] YouTube Help — “Your content & third-party training”
https://support.google.com/youtube/answer/15509945?hl=en
[S4] Meta — “Making AI Work Harder for Europeans”
https://about.fb.com/news/2025/04/making-ai-work-harder-for-europeans/
[S5] U.S. Copyright Office — “Copyright and Artificial Intelligence, Part 3: Generative AI Training”
https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf
[S6] European Commission — “General-purpose AI obligations under the AI Act”
https://digital-strategy.ec.europa.eu/en/factpages/general-purpose-ai-obligations-under-ai-act
[S7] SAG-AFTRA — “Artificial Intelligence”
https://www.sagaftra.org/contracts-industry-resources/member-resources/artificial-intelligence
[S8] SourceLess — STR Domains, Proof of Existence, and wNFT infrastructure
SourceLess Whitepaper v4.0
