Recent widespread adoption of generative artificial intelligence (GAI) tools has introduced a fundamental legal question: what cognizable right, if any, attaches to content produced by an AI system? As AI-generated text, images, music, source code, and other outputs become embedded in commercial workflows, the absence of a settled framework for ownership presents material risk for those who rely on such content as a commercial product or creative asset.

Because AI-generated outputs are embodied as text, audio, or other visual elements, many have turned to copyright law for resolution as to how ownership should be accorded to these outputs. The existing framework in copyright law, however, presents significant obstacles to the protection of AI-generated outputs: an AI-generated work is only eligible for copyright protection where there is meaningful, substantial input from a human author.

In the absence of certain protection for AI-generated works under copyright law, other legal theories, including contractual rights and common-law property torts, may offer partial alternatives. Even so, these other legal theories often fall short of providing the robust, nationwide exclusivity that copyright law affords. Consequently, many are left wondering how to best protect AI-generated content – at least until new guidance emerges on the protection of these assets.

Human Authorship: Copyright protection subsists in "original works of authorship fixed in any tangible medium of expression" under 17 U.S.C. § 102(a). Through decades of judicial construction and administrative practice, courts and the U.S. Copyright Office (USCO) have consistently held that "author" means a human being. A machine, algorithm, or AI system cannot be an author for purposes of the Copyright Act.

The Copyright Office's 2023 Policy Guidance: In March 2023, the USCO published a Statement of Policy entitled "Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence." In this statement, the USCO confirmed that works produced entirely by AI systems, without sufficient human creative control over the expressive elements, are not eligible for copyright registration. The USCO recognized, however, that human-AI collaborative works may qualify for limited copyright protection where a human author selects, arranges, or modifies AI-generated material in a sufficiently creative way. In other words, the AI-generated portions themselves remain unprotectable, and it is the applicant's burden to disclose and disclaim each of these AI-generated elements.

The D.C. Circuit's Decision in Thaler v. Perlmutter: On the heels of the USCO's issuance of the Statement of Policy, the U.S. District Court for the District of Columbia affirmed the USCO's decision to deny a claimant's copyright application for visual art generated by an AI system. On appeal, in March 2025, the Court of Appeals for the D.C. Circuit sided with the district court, holding that the Copyright Act requires all works to be authored by a human being.

The D.C. Circuit pointed to a number of provisions in the Copyright Act that frame various requirements in terms of human-only capabilities and make clear that authors must be human, including: the ownership provision, which assumes the author can hold property; duration provisions, which measure terms by the author's lifespan; the concept of joint authorship, which requires that multiple authors intend that their contributions be part of a unitary work; and the provision specifying that transferring copyright ownership requires a signature. For now, this decision stands, with the U.S. Supreme Court having denied the copyright claimant's writ of certiorari in March 2026.

Given the limitations under copyright law, parties have turned to contractual mechanisms to allocate ownership rights. Most commercial AI platforms include terms of service or enterprise license agreements that address ownership of outputs, with many expressly assigning to users all rights in generated outputs, to the extent any exist.

This contractual assignment conveys ownership to the user as between the user and the AI provider. However, this right arises from private agreement, not federal or state statute, and, as such, binds only the contracting parties. A third party who copies AI-generated output is not bound by the user's contractual ownership claim. Accordingly, without any underlying copyright in the original AI-generated output, there is likely no cause of action for infringement, no statutory damages, and no right to injunctive relief against third parties.

In the absence of enforceable copyright, remedies for misappropriation are substantially limited. Potentially viable theories may arise under state-based common law or statutory torts.