CREATOR Act
- Sponsor
- Rep. Van Duyne, Beth [R-TX-24]
- Committees
- Judiciary Committee (primary)
- Last action
- Jun 2, 2026
Bottom line
The CREATOR Act creates a novel federal IP right to protect visual artists from unauthorized AI-generated stylistic impersonation in commercial contexts, balancing artist protection with AI innovation and free expression.
What it actually does
This bill establishes a new federal intellectual property right for visual artists, granting them the exclusive authority to permit or prohibit the commercial exploitation or public distribution of a stylistic impersonation of their work, particularly when generated by artificial intelligence. It defines liability for unauthorized use, provides safe harbors for general-purpose AI systems and online services, and outlines specific exclusions for activities like parody, criticism, and news reporting.
Proponents argue
Proponents argue that the bill is essential to protect visual artists' livelihoods and creative control in an era where artificial intelligence can readily mimic their unique styles. They contend that existing copyright and trademark laws do not adequately address the challenge of AI-generated stylistic impersonation, and this new right provides artists with necessary legal recourse and economic agency over their artistic identity.
Opponents contend
Opponents might argue that the bill's definitions of 'stylistic impersonation' and 'distinctive visual characteristics' are overly broad and subjective, potentially stifling artistic influence, parody, and the legitimate development of AI technologies. They may also raise concerns about a chilling effect on free expression and innovation, as well as the potential for increased litigation and administrative burdens on online services and AI developers.
The bill is moderately complex due to new legal definitions and technical nuances; a careful reading is required to grasp its full implications.
Section 2(a)-(b)
Establishment of Exclusive Right Against Stylistic Impersonation
This provision creates a new federal intellectual property right for visual artists, giving them the exclusive authority to authorize the commercial exploitation or public distribution of a stylistic impersonation of their work. It clarifies that this right is distinct from copyright and trademark, does not extend to general ideas or artistic movements, applies only to deliberate impersonation, and can be licensed or assigned.
Supporters argue
Supporters argue this provision is crucial for protecting artists' control over their unique creative identity and commercial value in the face of advanced AI capabilities, ensuring they are compensated for the use of their style.
Critics contend
Critics contend that the definitions of 'stylistic impersonation' and 'distinctive visual characteristics' are inherently subjective, potentially leading to ambiguity, overreach, and a chilling effect on legitimate artistic influence and AI development.
Tradeoffs
This provision navigates the tension between providing robust protection for individual artists' creative output and avoiding undue restrictions on artistic inspiration, general AI development, and free expression.
Section 2(c)-(d)
Duration and Registration of Stylistic Impersonation Rights
For living visual artists, the right subsists for their lifetime. For deceased artists, it lasts 10 years after death, renewable in 5-year increments up to a maximum of 50 years post-mortem. Renewal requires filing a notice with the Register of Copyrights, which will maintain a publicly accessible online directory of these registered post-mortem rights.
Section 3(a)-(b)
Conditions for Liability and Protection for General-Purpose AI Systems
This section specifies that liability arises only if a person knowingly engages in commercial exploitation or public distribution of a stylistic impersonation without authorization. Crucially, it protects general-purpose AI systems from liability unless their provider intentionally configured the system to generate impersonations of a specifically identified visual artist AND expressly marketed it as such. Mere capability of an AI system to resemble an artist's work does not create liability.
Section 3(c)(2), Section 5
Online Service Provider Safe Harbors and Notice-and-Takedown Procedures
Online services are shielded from liability for user-generated content that constitutes stylistic impersonation, provided they have a designated agent, act expeditiously to remove or disable access upon receiving a valid notice of violation, and have a repeat violator policy. The bill outlines specific notification requirements, a counter-notification process for users, and establishes liability for those who knowingly misrepresent a claim of stylistic impersonation.
Section 4
Exclusions for Fair Use-like Activities
The bill explicitly excludes several activities from constituting a violation, provided they do not involve commercial exploitation intended to mislead as to source, sponsorship, or approval. These exclusions include commentary, criticism, scholarship, research, teaching, parody or satire (that critiques the artist), historical/biographical works, news reporting, and fleeting or incidental resemblances that do not reproduce a material combination of distinctive visual characteristics.
Section 6
Civil Actions, Injunctive Relief, and Statutory Damages
This section grants right holders the ability to bring civil actions against violators. Courts may award temporary or permanent injunctive relief. Prevailing plaintiffs can elect to receive either actual damages (plus the violator's profits) or statutory damages. Statutory damages range from $10,000 to $100,000 per work for intentional commercial impersonation, increasing to $50,000 to $150,000 for willful violations involving intentional targeting and commercial exploitation. Factors like willfulness, scale of dissemination, and commercial impact are considered for statutory damages.
Section 7, Section 8
Clarifications and Limited Preemption
These sections clarify that the Act does not limit or expand existing copyright law, grant rights over general ideas, impose monitoring duties on AI systems, or restrict First Amendment speech. It explicitly states that the mere capacity of an AI system to generate resembling outputs does not infer training data use, and civil actions cannot be based solely on training data claims. The Act preempts state laws only to the extent they impose liability for stylistic impersonation as defined herein, preserving other state and federal IP, right-of-publicity, and consumer protection laws.
Defines 'stylistic impersonation' as a visual work generated by an AI system that was intentionally configured, prompted, marketed, or designed to emulate an artist's distinctive visual characteristics AND reproduces them in a manner likely to mislead a reasonable viewer as to source, sponsorship, or approval, or to affect the commercial market for the artist's work. It also specifies what constitutes 'intentional design' and lists exclusions.
Section 10(11)
Why it matters:While placing definitions at the end is standard legislative practice, the detailed and nuanced nature of this particular definition means a casual reader might overlook its full implications for liability and scope.
Case for: Supporters argue that this precise and detailed definition is essential to narrow the scope of the new right, ensuring it targets only deliberate, misleading commercial exploitation by AI and prevents overreach that could stifle legitimate artistic influence or AI development.
Case against: Critics contend that terms like 'distinctive visual characteristics,' 'intentionally configured,' and 'likely to mislead a reasonable viewer' remain inherently subjective, potentially leading to extensive litigation, legal uncertainty, and a chilling effect on creativity and AI innovation.
Estimated impact: This definition dictates the entire scope of the new right and liability, profoundly impacting how visual artists are protected, how AI systems are developed and marketed, and how online content is moderated.
Defines 'specifically identified visual artist' as a visual artist whose name or distinctive visual characteristics are *expressly referenced* in the prompting, configuration, marketing, or operation of an artificial intelligence system, or in the promotion of the resulting output.
Section 10(10)
Why it matters:While definitions are typically at the end, the precise wording of 'expressly referenced' is a high bar that significantly influences how liability is established, making it easy to miss its full legal weight.
Case for: Proponents argue this specificity is vital to protect general-purpose AI systems and prevent liability for mere resemblance, focusing only on deliberate targeting. It aims to prevent frivolous lawsuits based on coincidental stylistic similarities.
Case against: Opponents might argue that requiring an 'express reference' sets too high a bar for artists to prove infringement, potentially allowing sophisticated AI systems to impersonate without direct naming, thereby circumventing the bill's protections.
Estimated impact: This definition significantly narrows the scope of liability for AI systems, impacting how AI is developed, marketed, and how artists can prove targeted impersonation.
States that the mere capacity of an artificial intelligence system to generate outputs resembling an artist's work does not create any presumption, inference, or evidentiary showing that the system was trained on that artist's work or dataset. Furthermore, no civil action under this Act may be predicated *solely* on claims regarding the training data, training process, or internal parameters of an AI system, unless such claims are independently actionable under other applicable law.
Section 7(b)-(c)
Why it matters:These are crucial limitations on how the law can be applied, particularly relevant to ongoing, highly contentious debates about AI training data and intellectual property, and their placement here provides significant protection for AI developers.
Case for: Supporters argue this protects AI developers from liability based solely on the resemblance of outputs, which could be coincidental or due to general stylistic influence, rather than direct copying in training. It aims to prevent speculative lawsuits over AI training data and processes.
Case against: Critics might argue this provision makes it significantly harder for artists to prove that their work was used without authorization in AI training, even if the resulting output is clearly imitative, potentially undermining other IP rights and shifting the burden of proof unfairly.
Estimated impact: This provision provides substantial legal protection for AI developers regarding their training data and processes, potentially influencing the landscape of intellectual property litigation related to AI outputs.