I came across an article on LinkedIn recently about AI authorship. It asked whether works created with significant AI assistance should receive the same copyright protection as traditionally authored works. The question stayed with me. The more I considered it, the less confident I became that we are approaching the problem correctly.
The concern behind the question is legitimate. If someone enters a short prompt, receives a 70,000-word novel, makes a few cosmetic changes, and publishes it under their name, there is a real question about authorship.
But we need to look deeper than that. Somewhere in this conversation, we began treating "AI-assisted" as though it describes a single method of creation. It does not. If we build copyright rules around that assumption, those rules will make less and less sense as AI becomes embedded in the tools we use to think, write, research, and communicate.
The question is not whether AI was used. The harder and more useful question is where the meaningful human creative agency resided.
We Do Not Own the Language
Copyright protects original expression. It does not give us ownership of ideas, facts, concepts, methods, individual words, or the language itself.
Every sentence in this article is built from words I did not create. I am selecting them, arranging them, and using them to communicate ideas. Consider this sentence: "The evidence does not establish who accessed the system." Can I own that arrangement of words because I happened to write it? Someone has almost certainly written something close to it. Possibly the exact words. There are only so many sensible ways to express a given idea.
That protection has limits. Not every arrangement of ordinary language receives the same scope of protection. If common combinations of words could be removed from use because someone had written them before, communication would eventually become impossible. Copyright law recognizes this tension. Protecting expression cannot become ownership of the underlying idea or of the language needed to communicate it.
That distinction matters when we introduce generative AI.
What Happens When AI Improves My Sentence?
Suppose I develop an argument about AI-assisted investigations. I write:
"Organizations using AI for investigations need to keep enough records so that someone can later understand how the AI contributed to the conclusion."
The thought is mine. The argument is mine. But I do not like the sentence. So I ask an AI system to make it more concise without changing the argument. It returns:
"Organizations deploying AI within investigative workflows should maintain sufficient records to enable subsequent reconstruction of AI-supported conclusions."
Now I have a question. At what point did I stop being the author?
The proposition originated with me. I decided what to communicate. I provided the content. I directed the transformation. I evaluated the result against my own thinking. I decided whether those words belonged in my work.
What ran through every step was judgment. I decided what the idea was, and whether the revised sentence still carried it. The AI produced the language. The judgment about that language was mine. That exercise of creative judgment should matter when we assess authorship.
Now replace the AI with a professional editor. I send the editor my original sentence and say I do not like how I expressed it. The editor returns the same revised sentence. Would we conclude that I had ceased to be an author? I do not think we would.
This is where the distinction becomes uncomfortable. If functionally similar assistance produces opposite conclusions about authorship because one assistant is human and the other is computational, then the test itself deserves examination.
The Tool Cannot Be the Test
Writers have never created in isolation. We use spellcheckers, grammar correction, predictive text, search engines, reference databases, transcription software, translation tools, researchers, and editors. Nobody argues that using these tools eliminates authorship.
Generative AI extends this line, sometimes dramatically. But the presence of AI tells us little about who created the work.
Consider two authors. Author A writes, without assistance:
"As AI becomes embedded in investigations, investigators may struggle to reconstruct and explain how their conclusions were reached."
Author B develops the same idea but expresses it poorly at first, then asks an AI system to improve the wording. The AI happens to produce the same sentence.
The words on the page are identical. One was composed directly by the human author. The other emerged through AI-assisted revision. In both cases the author decided what to say and judged whether the final sentence said it. The only difference is how the words were produced. If we treat one as protectable and the other as automatically excluded, we have to ask what we are protecting. Are we protecting human creativity, or the provenance of keystrokes? Those are not the same thing.
AI Assistance Is Not AI Authorship
There is another side to this, and it matters as much. I am not arguing that everything produced with AI qualifies as human-authored work. That would replace one overly broad position with another.
Suppose someone types: "Write me a 70,000-word political thriller about an investigator uncovering corruption inside an AI company." The system develops the characters, the dialogue, the scenes, the narrative, and nearly all of the prose. The person makes a few minor corrections and publishes it. There is a legitimate question about what that person actually authored. Having an idea is not the same as authoring its expression.
But notice where the problem sits. It is not that AI was used. It is that the human creative contribution may be insufficient to support a claim of authorship over substantial portions of the expression. That is a different test, and a better one.
There Is a Continuum We Cannot Ignore
Consider the range:
- A writer uses AI for spelling and grammar corrections.
- A writer asks AI to suggest clearer sentences.
- A writer asks AI to restructure paragraphs they have already written.
- A writer develops an argument and uses AI to challenge it and expose weaknesses in the reasoning.
- A writer debates an idea with an AI system, rejects some suggestions, accepts others, and develops their thinking through the exchange.
- A writer produces part of a work while AI generates other parts, then extensively directs, modifies, and selects the generated material.
- A writer provides a short prompt and accepts an entire generated manuscript.
These are not the same creative process. So where did the human stop being an author? At the first AI-generated word? The first rewritten sentence? Twenty percent of the manuscript? Fifty?
There is no satisfying answer if the starting point is only "AI was involved." The category is too crude for the problem.
The Editor Comparison Has Limits
The comparison with a human editor is useful, but I do not want to overstate it. Editors can make original creative contributions. An editor who substantially rewrites a work can raise real questions about authorship and ownership. So my argument is not that AI is merely another editor and nothing has changed.
Something has changed. Generative AI can contribute to creative work at a scale earlier tools could not. That deserves serious attention. But the principle underneath our response should hold: authorship depends on meaningful human creative contribution, not on which tools participated in producing the final work.
Otherwise We Are Measuring Purity, Not Creativity
Imagine someone spends three years developing an original theory. They conduct the research, develop the methodology, analyze the evidence, construct the argument, challenge their assumptions, and determine the conclusions. Writing is not their strongest skill, so they use AI extensively to communicate those ideas clearly.
Now imagine someone else writes a conventional but independently authored 200-page manuscript without touching an AI system.
Is the second person more of an author? If our answer is yes because they did not use AI, then we have stopped measuring creativity. We are measuring technological purity. I am not convinced that is what copyright was designed to protect.
Publishing Does Need Standards
None of this means the industry should ignore the problem. We need better language for describing human-AI creative relationships. "AI-generated" and "AI-assisted" cannot keep carrying all the weight.
There is a meaningful difference between a work substantially generated by a system with minimal human intervention and one where AI helps a human develop, challenge, edit, or communicate substantially human-created ideas. There may be useful distinctions between AI-generated, AI-directed, and AI-assisted work. Disclosure may matter, particularly where AI has produced substantial expressive portions of a work.
Those are conversations about provenance, contribution, and creative control. They are more useful than asking whether AI touched the work.
We May Be Asking the Wrong Question
This is what stayed with me after reading the article. The publishing industry is right to take AI authorship seriously. But we may be trying to solve a twenty-first-century authorship problem with an increasingly binary idea of creativity. Human or machine. Authored or generated. AI or no AI.
Creativity does not work that way, and it may never have. We build on language we inherited. We develop ideas through conversation. Editors challenge our writing. Researchers help us find information. Technology helps us communicate. AI has now entered that ecosystem. That creates new problems, and some of them are serious. But a new tool should not lead us to confuse assistance with authorship, or authorship with typing.
So the question for publishers, authors, lawyers, and policymakers should not be whether AI was used. The better questions are harder. What did the human actually create? What substantive creative decisions did the human make? What did the machine contribute? Who exercised meaningful control over the final work? And where machine-generated expression exists, can we separate it from the human-authored work around it?
Those questions are more honest. If our answer to AI is a line between work that has touched it and work that has not, that line will become impossible to defend. Eventually, it will become impossible to find.
The future of authorship should not be decided by whether a machine helped us find the words. It should be decided by whether the ideas, choices, judgment, and creative agency behind those words remained meaningfully human.