ThatPainter is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn More
More than 10,500 creators, creative-industry organizations, and cultural institutions initially signed the “Statement on AI training”, published on October 22, 2024. The short statement argues that using creative works without a license to train generative-AI systems threatens the livelihoods of the people who make those works.
Although headlines often call the signers “artists,” the coalition extended well beyond visual art. Actors, musicians, authors, photographers, composers, publishers, and rights organizations were among those represented. The statement was a public-policy demand—not a court ruling—and it did not establish that every use of copyrighted material for AI training is illegal.
What the open letter said
The statement’s central message was deliberately brief:
#1 Best Overall
“The unlicensed use of creative works for training generative AI is a major, unjust threat to the livelihoods of the people behind those works, and must not be permitted.”
The quotation appears on the statement’s official website. It expresses a broad principle: companies developing generative-AI systems should not use creative works for training without authorization or licensing.
The letter did not name one defendant, identify a particular dataset, propose a single licensing system, or request damages. It also did not provide a technical definition covering every form of web scraping, text-and-data mining, public-domain material, licensed datasets, or user-uploaded content. Its force came from the breadth of the coalition and the policy position it presented, rather than from detailed legal or technical provisions.
Who signed it?
The initial signatories included individuals and organizations from several creative sectors. Reported examples included:
- Actors: Julianne Moore, Kevin Bacon, Rosario Dawson, F. Murray Abraham, Kate McKinnon, and Sean Astin.
- Musicians and composers: Thom Yorke, Björn Ulvaeus, Robert Smith, Billy Bragg, Max Richter, Kate Bush, and Geoff Barrow.
- Authors: Kazuo Ishiguro, James Patterson, Ian Rankin, Malorie Blackman, William Boyd, and Tracy Chevalier.
- Organizations: the International Federation of the Phonographic Industry, News/Media Alliance, publishing and authors’ groups, and Penguin Random House.
The names are examples, not the substance of the campaign. The official signatory page is the best source for the live list.
The initial announcement on October 22 described more than 10,500 signatories. The website later displayed 50,544 signatories, according to the supplied source material. Those figures describe different points in time: the larger number should not be presented as the count at launch, and neither number should automatically be read as the number of individual visual artists. The total includes organizations and participants from many creative professions.
Who organized the statement?
The campaign was associated with Ed Newton-Rex, a former Stability AI executive who later founded the nonprofit Fairly Trained. Publishers’ Licensing Services said Newton-Rex resigned from Stability AI in 2023 over concerns about the use of copyright-protected works without permission. The organization later joined the international statement.
Support came from groups representing publishing, music, photography, authors, news media, and other creative fields. The IFPI described itself as one of the initial signatories, while the News/Media Alliance reported its support alongside the initial signatory figure.
Free tools Windows power users keep installed
One-click scans. No signup required.
Why are painters and other creators concerned?
The signatories’ objections center on control over creative work and the economics of making it. Their concerns include:
- Consent: A creator may not have agreed to have paintings, photographs, writing, music, performances, or other work included in training data.
- Compensation: Creators argue that commercial AI companies can derive value from their work without paying the people who produced it.
- Competition: Generative systems can produce images, text, music, voices, and performances that may compete with human-made work or reduce demand for commissioned work.
- Attribution and transparency: Creators may have no reliable way to determine whether a particular work was used or how it influenced a model.
- Livelihoods: The statement characterizes unlicensed training as a threat to the economic position of creative workers.
- Cultural production: Publishers, labels, unions, and creators argue that large-scale unlicensed copying could weaken the industries that finance new creative work.
These are the signatories’ concerns and policy arguments, not proof that every AI system has produced the same economic effect or that every creator’s work was used unlawfully.
Rank #3
What does “unlicensed” mean?
In practical terms, licensed training means that an AI developer has obtained permission or rights to use specified works, generally through a contract or dataset license. Unlicensed training means the developer used material without obtaining that permission from the relevant rights holder.
That distinction is not the same as a final legal judgment. A work being publicly accessible online does not automatically mean it is freely licensed for commercial AI training. At the same time, whether copying material for training is lawful can depend on the jurisdiction, the source material, the purpose and manner of copying, applicable exceptions, and the facts of a particular case.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →An opt-out system is also different from prior permission. An opt-out gives a creator a way to request exclusion after—or sometimes before—use. It does not necessarily provide consent, payment, or knowledge of what was already included in a dataset. The letter’s position was that commercial use of creative works for training should require authorization rather than treating online accessibility as blanket permission.
The legal question is still unsettled
The protest took place amid copyright lawsuits involving AI developers and wider disputes over whether training may qualify as fair use, fair dealing, text-and-data mining, or another legal exception. Those questions are jurisdiction-specific and fact-specific; the statement did not resolve them.
It is also important to separate several disputes that are often bundled together:
- Training data: whether works may be copied into datasets or used to train a model.
- Generated outputs: whether a particular image, passage, song, or other output infringes a protected work.
- Style: whether an output resembles a living artist’s style. Copyright in a specific work is not the same as a general legal right to control an artistic style.
- Employment and performance rights: whether actors, writers, musicians, or other workers consented to digital replicas, voice cloning, or AI-generated performances under a contract or collective-bargaining agreement.
In that sense, the signatories were trying to shift the debate from whether companies could technically collect material available online to whether commercial model developers should be required to obtain permission and pay for creative works used in training. That is a policy interpretation of the campaign, not a court’s conclusion.
What policy changes did supporters want?
The strongest documented demand was licensing. The Authors Guild characterized the campaign as a call for regulators to require AI companies to license the creative works on which they train.
That goal could be implemented in different ways, including direct contracts, collective licensing, dataset disclosure, enforceable opt-out systems, or consent and opt-in requirements. Supporters have also raised related issues such as payment, protection against unauthorized voice and likeness replication, labeling of synthetic content, and contractual safeguards for workers. However, those are broader policy proposals; they were not all spelled out in the statement’s single sentence.
What did publishers and other organizations do?
Penguin Random House said it joined the coalition and opposed the unauthorized use of copyrighted content to train generative-AI models. The publisher also said it began adding a copyright-page notice stating that its books may not be used for AI training. Its announcement made clear that this was the publisher’s position.
A notice on a copyright page can document a rights holder’s objection and potentially support later contractual or legal arguments. It does not, by itself, guarantee that every AI system will comply, or establish that every author’s contract has identical terms. Nor does a printed notice alone settle the legal status of every possible training use.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsFairly Trained represents another approach: its stated certification program addresses whether an AI company uses licensed training data. That is narrower than a general certification that a model is ethical, safe, accurate, or noninfringing. Its focus is the claimed licensing of training data, as described by the organization.
How was this different from entertainment-industry strikes?
The statement was not a union strike or collective-bargaining action. It overlapped with broader concerns raised during entertainment-industry labor disputes, including consent and payment for AI-generated performances, digital replicas, and voice likenesses.
Those negotiations primarily concern employment relationships, contracts, and union protections. The open letter addressed training material across a much wider set of industries, including visual art, publishing, music, photography, and acting. A performer’s contractual right to approve a digital replica is a different question from whether a company could use a copyrighted painting, photograph, book, or recording in a training dataset.
Did the letter change anything?
Its immediate effect was political, reputational, and organizational. The signatory count showed public opposition from a large, cross-industry coalition, while institutional support connected individual creators to publishers, labels, media organizations, and rights groups.
Recommended Free Tools
But the statement did not ban AI training, stop a named company, create an enforceable licensing requirement, or determine the outcome of a copyright case. Its practical effect depends on what follows: licensing markets, dataset transparency, legislation and regulation, litigation, publisher policies, contracts, and collective action.
The central unresolved question is not whether creators object. It is how permission, compensation, transparency, and liability should work in practice. The October 2024 statement made the coalition’s preferred starting point clear: creative works should not be used to train generative-AI systems without a license.
Quick Recap
