Legally Speaking, Who Owns What AI Creates?

Morey J. Haber, Chief Security Advisor at BeyondTrust, is an identity and technical evangelist with over 25 years of IT industry experience.

getty​Generative artificial intelligence has created a problem technology has encountered many times before: We invent new capabilities before we completely figure out the rules.

We can ask generative AI to produce an image, compose music, write software or design a business plan in seconds. With this, adoption has moved from novelty to default. Seventy-two percent of organizations report using generative AI in at least one business function, nearly double the share two years earlier, according to McKinsey’s 2025 State of AI survey.

Now, the legal system is reminding us about a deceptively simple question. Who owns the content? The answer is considerably more complicated than clicking “accept” on a licensing agreement or acceptable use policy.​

For most of modern copyright history, ownership has started with a human creator. Copyright law then provided a framework for determining ownership, licensing, reproduction and compensation.

Generative AI complicates this model because the person requesting the content may contribute only a prompt, while the machine performs much of the creative work.

Consider a simple prompt: “Create a photograph of a futuristic city at sunset.” The user supplied the idea, but ideas are generally not protected by copyright. Behind the scenes, embedded in technology, the AI system interpreted those instructions, the software provider supplied the technology, and the model developers created the algorithms. The training process may have incorporated enormous quantities of material originally created by other people. Therefore, the output appears to be new and unique, but is actually built from the intellectual property of countless human creators.

Suddenly, a simple image has a complicated ancestry and raises a difficult question: Who is the author? In the United States, the emerging answer begins with an important principle. Copyright protection requires human authorship.

The U.S. Copyright Office reaffirmed that standard in its January 2025 report on AI and copyrightability, following a 2023 decision denying copyright for the AI-generated illustrations in the graphic novel Zarya of the Dawn while upholding copyright in the human-authored text and arrangement. If a person merely enters a basic prompt and accepts the resulting output, there may not be enough human creative contribution to establish copyright in that output.

That does not mean AI-assisted works cannot receive copyright protection. If an author uses AI to generate material and then substantially selects, arranges, rewrites, edits or transforms that material, those human contributions may qualify for protection. The distinction is increasingly important. Using AI as a tool is different from asking AI to be the creator.

No one would argue that using Photoshop to adjust a photograph makes the software its author. The photographer remains responsible for the creative decisions. Generative AI complicates that relationship because it can generate substantive creative elements the human did not explicitly design.​

Then there is a second question. What rights does the AI provider have?

AI providers can contractually define rights concerning inputs, outputs, data usage, licensing and commercial use. But those rights are separate from whether copyright law recognizes an output as copyrightable. A platform’s terms of service cannot create copyright protection where the law does not recognize it.

Training data adds another layer. Generative AI models learn from enormous datasets containing books, photographs, websites, software, music, artwork and other material, some of which may be copyrighted. Courts, regulators, creators and technology companies continue to grapple with when training constitutes permissible use, when licensing is required, and whether outputs can infringe on existing works.

The distinction businesses cannot afford to overlook is that having permission from an AI provider to use an output does not necessarily mean the output is free from third-party intellectual property or copyright risk.

Ask AI to create a corporate mascot that looks suspiciously similar to an existing copyrighted character, for example, and you may have permission under the AI provider’s terms to use the generated image while simultaneously creating an intellectual property issue with somebody else.​

AI-generated software creates another interesting scenario. Developers increasingly use generative AI assistants to produce functions, scripts and application components. As of 2025, 84% of developers reported using AI coding tools, according to Stack Overflow’s Developer Survey. If generated code reproduces protected material or introduces unexpected licensing obligations, organizations may inherit legal, compliance and software supply chain risks they never intended to accept.

But software also exposes the next evolution of this problem. There is a meaningful difference between AI generating code for a human to review and AI being given the ability to act on that work. As AI systems become more agentic, they may be authorized to access repositories, modify files, invoke tools, execute commands or initiate workflows on a user’s or organization’s behalf.

At that point, the question expands beyond who owns what AI created. Organizations also need to consider who authorized what AI was allowed to do, where its authority came from, and who is responsible for the result.

Enterprises therefore need policies defining which AI systems employees may use, what information may be submitted, how generated material can be used commercially, and when human review is required.​

There is a fascinating irony emerging from all of this. Companies are racing to use AI because it can create content faster and cheaper than humans. But the more autonomous that creation becomes, the more complicated ownership can become. Human involvement may ultimately become one of the strongest arguments for establishing copyright claims around AI-generated content.

Still, ownership may be only the beginning. Reducing direct human involvement in execution does not remove the organization’s responsibility. If anything, greater AI autonomy makes it more important to establish a clear line between what the technology can do and the people accountable for deciding what it should do.

The question today may be who owns what AI creates. The question tomorrow will be who owns what AI does.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?