Mind the Gap: States Step in to Address Ownership of AI-Generated Works Amid Federal Uncertainty

Following the Supreme Court’s denial of certiorari this spring in Thaler v. Perlmutter (Case No. 25-449), which left unaddressed the U.S. Copyright Office’s (“USCO”) requirement of human authorship for copyright protection, creators are left without clear guidance on just how much human contribution, if any, is required for a work to be copyrightable.

Enterprising legislators may rush to fill this void, and Arkansas is ahead of the game, having passed a law last year that establishes an ownership right—separate from the ownership rights created by federal IP law—for the outputs of AI systems.¹ With exceptions for generative works created as part of a person’s employment, which are treated akin to a “work for hire,” the law grants “ownership” rights to “the person who provides the input or directive to the generative artificial intelligence tool . . . provided that the content does not infringe on existing copyrights or intellectual property rights.”² The law also clarifies that the “person who provides data or input to train a generative artificial intelligence model shall be the owner of the resulting trained model,” as long as the data used to train the model “is lawfully acquired” and there has not been a transfer of ownership.³ The law does not, however, provide for any specific rights or remedies, leaving open questions as to how these ownership interests may be enforced. (Courts might apply traditional property law remedies available under Arkansas common law, including claims for conversion or unjust enrichment.)

This year, Iowa lawmakers proposed similar bills in both the State House⁴ and the State Senate.⁵ The proposed bills closely track the language of the Arkansas law passed last year, and similarly include language expressly separating the ownership rights granted through the state law from any existing federal copyright or patent rights. Neither version of the Iowa bill made it out of committee prior to the close of the legislative session at the beginning of May.

While gaps in federal law leave room for the states to step up and establish clearer guidelines regarding the ownership of AI outputs, these laws risk running afoul of preemption. The preemption provision of the Copyright Act provides:

On and after January 1, 1978, all legal or equitable rights that are equivalent to any of the exclusive rights within the general scope of copyright as specified by section 106 in works of authorship that are fixed in a tangible medium of expression and come within the subject matter of copyright as specified by sections 102 and 103, whether created before or after that date and whether published or unpublished, are governed exclusively by this title. Thereafter, no person is entitled to any such right or equivalent right in any such work under the common law or statutes of any State.⁶

Federal courts have established a two-pronged test, holding that a state law is preempted (1) if the intellectual property at issue is within the “subject matter of copyright,” and (2) if the claimed property rights are “equivalent to” the federal rights granted through copyright law.⁷ While there is a wealth of precedent regarding preemption and the overlap between copyright law and other related rights, such as the right of publicity, there is less clarity as to how preemption applies in the context of emerging technologies such as AI—particularly in areas where the property in question has been, so far, expressly excluded from federal copyright protection.

Although neither the Arkansas law nor the proposed bill in Iowa claims to establish copyright law or circumvent existing federal intellectual property laws, they do govern ownership rights directly adjacent to these federal protections; as such, it would be unsurprising for these laws to face challenges on preemption grounds—though the likelihood of success of such a challenge remains uncertain. If relevant state laws remain within an area explicitly outside the bounds of federal copyright law—ownership of fully AI-generated works that are expressly not copyrightable—there may be sufficient room to overcome preemption arguments, and to offer a degree of clarity and protection to users of AI models.

Preemption is not the only challenge facing states’ attempts to govern ownership of AI outputs. There is also the potential that mismatched rights across jurisdictions ultimately lead to more confusion and uncertainty. A state law may struggle to offer businesses and individual creators meaningful confidence in their rights associated with AI-generated content if a neighboring state offers differing standards for ownership or mechanisms for relief; this is particularly true for those who work across state lines. These inconsistencies could ultimately create greater uncertainty regarding who—if anyone—owns AI-generated content.⁸

States’ efforts to address ownership of AI-generated works underscores the need for a coherent federal framework. Until Congress or the courts articulate a clear standard for the degree of human involvement necessary to warrant intellectual property protection—or, in the alternative, a distinct ownership regime for AI-generated works—creators and businesses should continue to monitor developments on both a federal and state level in order to best understand their rights and protect their works.

Thaler, which involved a work that was fully generated by AI with no claimed human authorship, served as a test at the far end of the spectrum and ultimately left in place the standard that at least some human authorship is necessary for a work to be protected by copyright. Pending cases, such as Allen v. Perlmutter,⁹ in which the plaintiff is challenging the USCO’s refusal to register a work generated through hundreds of prompts directed and refined by a human author, may ultimately establish clearer guidance through the federal court system on where the line falls between protectable human work and AI-generated material not subject to copyright protection. However, in the interim, certain states have entered the ring, potentially offering an alternate avenue through which creators and businesses who utilize AI tools in their work can gain clarity.

¹ Ark. Code Ann. § 18-4-101 (2025).

² Id.

³ Id.

⁴ H.F. 2691, 91st Gen. Assemb. (Iowa 2026).

⁵S.F. 2199, 91st Gen. Assemb. (Iowa 2026).

⁶ 17 U.S.C. § 301(a).

⁷ See Crow v. Wainwright, 720 F.2d 1224, 1225-26 (11th Cir. 1983), cert. denied, 469 U.S. 819 (1984).

⁸ For additional commentary on the commercial and business implications of the uncertainty surrounding ownership of AI-generated content, see Jacob Noti-Victor and Xiyin Tang, The Secret Weapon Against AI Dominance, The Atlantic, April 30, 2026.

⁹ Allen v. Perlmutter, No. 1:24-cv-2665 (D. Colo. Sept. 26, 2024).