The question of if AI-generated images, illustrations, and text can be protected by copyright is one that editors and publishers are confronting with increasing urgency. The short answer, under current US law, is no: purely AI-generated works cannot be copyrighted. But the fuller picture is more layered, and understanding it matters for every newsroom experimenting with AI-assisted production.

The Core Legal Position

The US Copyright Office has maintained a consistent position: works that lack human authorship cannot be registered. This principle is not new. Copyright law has long required that a human being be the creative force behind a work. What is new is the pressure that AI image generators and large language models are placing on that principle, forcing regulators and courts to spell out exactly what it means in practice.

Two cases have become the clearest reference points. In the matter of Thaler v. Perlmutter, Stephen Thaler submitted a work called "A Recent Entrance to Paradise," listing an AI system as the sole author. The Office denied registration on the grounds that the image lacked human authorship. Thaler challenged the decision, but courts have upheld the Office's reasoning.

The second case involved the graphic novel "Zarya of the Dawn," created by Kristina Kashtanova using Midjourney-generated images. The Copyright Office's ruling here was instructive: Kashtanova retained copyright over the human-authored elements of the work, specifically the text and the creative arrangement of the pages, but the AI-generated images themselves were excluded from protection. The Office found that she had not exercised sufficient creative control over the individual images to qualify as their author.

What "Sufficient Human Authorship" Actually Means

The Zarya ruling points to the practical question that newsrooms and visual teams need to think about carefully: at what point does a human's involvement in an AI-assisted creative process cross the threshold into authorship? The Copyright Office has indicated it will evaluate this case by case, looking at factors such as:

  • The degree to which a human selected, arranged, or modified AI output
  • Evidence that the human made creative choices that shaped the final work
  • The extent to which the AI output was transformed through human editing

Writing a detailed prompt and accepting the first output the model produces is unlikely to establish authorship. Substantially editing an AI-generated image, making deliberate compositional choices, or combining AI output with original human-created elements may support a stronger claim. The line is genuinely uncertain, and the Office has not yet issued comprehensive guidance that resolves every scenario.

This matters directly for the broader copyright fight between publishers and AI companies, because the question of who owns AI-assisted content sits at the heart of how newsrooms can protect their output.

The Editorial Angle: What This Means for Newsrooms

For publishers, the practical implications run in two directions. First, newsrooms that use AI tools to generate images or text need to understand that those outputs, if not substantially shaped by human creative choices, may not be protectable. That affects licensing, syndication, and the ability to pursue infringement claims if a competitor copies the work.

Second, and perhaps counterintuitively, the same legal framework that denies copyright to AI-generated work also complicates the position of AI companies themselves. If an AI system cannot hold copyright, questions arise about the status of works that AI models produce from training data drawn from copyrighted human work. Our coverage of the landmark case between the New York Times and OpenAI and the running tracker of AI copyright lawsuits in 2026 shows how actively courts are working through the downstream consequences of these unresolved questions.

Some publishers have moved to address the uncertainty through commercial agreements rather than waiting for legal clarity. AI licensing deals between publishers and LLM companies represent one attempt to create contractual protections where copyright law does not yet provide them automatically.

What Editors Should Do Now

Given the current state of the law, there are several concrete steps editorial and legal teams can take:

  • Document human creative choices made during any AI-assisted production process, including prompts, editing decisions, and modifications to outputs
  • Treat purely AI-generated images as unprotected unless a human author has demonstrably shaped them
  • Review contracts with contributors who use AI tools, and clarify ownership expectations explicitly
  • Consult legal counsel before attempting to register AI-assisted works with the Copyright Office
  • Monitor Copyright Office guidance, which is evolving and may be updated as cases accumulate

The law is not static. Congress has received recommendations and testimony on AI and copyright, and the Copyright Office issued a report in 2023 opening a formal study of the area. Legislative movement is possible, though timing remains unpredictable.

Sources

  • US Copyright Office, Zarya of the Dawn registration decision (February 2023)
  • US Copyright Office, refusal of Stephen Thaler's "A Recent Entrance to Paradise" application
  • US Copyright Office, Copyright and Artificial Intelligence study (initiated 2023)
  • Thaler v. Perlmutter, US Court of Appeals for the Federal Circuit (2022)