Policy
AI-Generated Work May Have No Copyright Owner, Creating Legal Gap
Courts worldwide are reaching conflicting conclusions on whether machine-generated content can be protected, leaving employers uncertain about ownership of employee deliverables.
Key takeaways
- US courts require human authorship for copyright protection, meaning purely AI-generated content likely has no owner under current law.
- The same AI-generated work may be protected in China, unprotected in the US, and protected temporarily in the UK under a provision the government wants to remove.
- Employees entering confidential information into public AI tools may destroy privilege protections, as one federal court ruled such exchanges are not confidential.
- Employment contracts can immediately address ownership gaps by specifying approved tools, requiring documentation of AI contributions, and assigning all potential rights to employers.
- Firms cannot accurately answer clients about deliverable ownership unless they tracked which portions were human-created versus machine-generated during production.
When a law firm associate uses an AI tool to draft a shareholder agreement and then refines it, who owns the result? The answer depends on which country's law applies—and for the portions generated purely by the machine, the answer in several major jurisdictions may be that nobody owns it at all.
This uncertainty stems from a fundamental mismatch between traditional copyright law and generative AI. Employer ownership rules were designed around a simple premise: a human employee creates work, and the law determines whether the creator or employer holds the rights. But if no human authored the initial output, those ownership rules have nothing to transfer.
Diverging legal standards
The United States has taken the clearest position requiring human authorship. In Thaler v. Perlmutter, the US Court of Appeals for the DC Circuit held in March 2025 that copyright law requires a human author, a decision the Supreme Court declined to review in March 2026. The Copyright Office's January 2025 report went further, concluding that prompts alone do not give users enough control to qualify as authors, though human selection, arrangement, and modification can still receive protection.
For the associate's shareholder agreement, this means her edits are likely protected while the machine's initial draft probably is not. Her firm holds rights to the editing but has a questionable claim to the underlying text.
The United Kingdom took a different approach in 1988, granting copyright in computer-generated works to whoever made the arrangements necessary for creation. However, the UK government's March 2026 report recommended removing this provision after consultation respondents overwhelmingly opposed protecting purely AI-generated works. Firms relying on this protection are depending on a statute their own government wants to eliminate.
China presents yet another model. The Beijing Internet Court ruled in November 2023 that an image created with Stable Diffusion warranted protection because the user's choices reflected intellectual input. The same machine-generated paragraph could therefore be protected in Beijing, unprotected in Washington, and protected in London temporarily.
Why it matters
This jurisdictional patchwork creates immediate business risks beyond theoretical ownership questions. When Samsung engineers reportedly pasted internal code into ChatGPT in early 2023, the company banned generative AI tools within weeks. In February 2026, Judge Jed Rakoff ruled in United States v. Heppner that exchanges with a consumer AI platform were not covered by attorney-client privilege because the platform's terms allowed disclosure of user inputs. An employee entering client material into a public AI tool may inadvertently destroy confidentiality protections that existing contracts never contemplated.
Liability poses another concern. In Mata v. Avianca (2023), a federal judge sanctioned both individual lawyers and their firm $5,000 after they filed a brief citing AI-invented cases. The firm was held responsible for work it never explicitly approved.
Contractual solutions
While courts and legislatures will take years to resolve these questions, employment contracts can address the gap immediately. Four provisions would close most exposure: specifying which AI systems employees may use and on what account type; prohibiting entry of confidential information into unapproved systems; requiring employees to assign all rights in AI-assisted work to the employer and to document which portions were machine-generated; and designating who must verify facts and citations before any AI-assisted work is delivered.
The documentation requirement serves a practical purpose beyond ownership. The US Copyright Office requires applicants to disclose AI-generated content and disclaim any that is more than minimal. Employers who cannot trace the machine's contribution cannot make that disclosure accurately.
Clients asking who owns delivered work deserve an honest answer: the human contribution is protected, the machine's contribution may not be depending on jurisdiction, and firms can only distinguish between the two if they recorded the difference during creation. Until the law settles, employment contracts are determining ownership by default.
These details were first reported by Sheraz Ahsan writing for JURIST.