On the Impossibility of Owning Coordinates: Why Prompts Cannot Be Proprietary

 J. Rogers, SE Ohio

Abstract

This paper examines claims of ownership over AI prompts and argues that such claims are fundamentally incoherent both philosophically and legally. Drawing parallels to navigational directions and established copyright doctrine, we demonstrate that prompts are coordinates in a collectively-created conceptual space rather than protectable works of authorship. Just as one cannot own a gas station by virtue of giving directions to it, one cannot own regions of semantic space by virtue of describing a path to them. More critically, prompts fail to meet basic legal requirements for copyright protection: they are functional instructions, not expressive works; they are too brief and utilitarian; and they consist primarily of unoriginal, standard communicative patterns that fall under the scènes à faire and merger doctrines.

Introduction: The Misdirection of Ownership

Imagine someone claiming ownership of a gas station because they gave you directions to reach it. The absurdity is immediate. The gas station existed before the directions. The roads existed before the directions. The cardinal directions, the street names, the landmarks—all preceded any individual's description of how to navigate them. The directions are merely a description of a path through shared infrastructure.

Yet this is precisely the claim made by those asserting ownership over AI prompts. They mistake the map for the territory, the coordinates for the destination, the description for the thing described. More fundamentally, they mistake a functional command for a copyrightable work of authorship.

The Nature of the Conceptual Commons

Language Models as Collective Artifacts

Large language models are trained on vast corpora of human writing—books, articles, conversations, code, poetry. They are statistical compressions of humanity's collective expression. The semantic space they encode is not created by the model, nor by any individual prompter, but emerges from millions of humans contributing to the linguistic commons over generations.

When you prompt a language model, you are navigating a space that was built by:

• Every author whose work was in the training data
• Every speaker of the language whose usage patterns shaped grammar and meaning
• Every domain expert whose technical writing established terminology
• Every creative writer who expanded the boundaries of expression

This is not mere metaphor. The model's "understanding" of concepts exists only because countless humans used language in particular ways. The conceptual space is literally constituted by collective human activity.

Prompts as Functional Instructions, Not Works of Authorship

A prompt does not create anything new in the model's conceptual space. It specifies a location within that space, a direction of travel, a set of constraints on which region to explore. Consider:

Prompt: "Write a sonnet about artificial intelligence in the style of Shakespeare"

This prompt coordinates several pre-existing regions:

• The concept "sonnet" (established by centuries of poetic tradition)
• The domain "artificial intelligence" (established by decades of technical and popular writing)
• The style "Shakespeare" (established by one author's corpus and centuries of literary analysis)

The prompt is a vector in semantic space. It points. It does not create.

Crucially, prompts lack the requisite "original work of authorship" required for copyright protection. They are imperatives and coordinates, not narratives or expressive compilations. They are closer to a recipe's instructions ("fold in the egg whites") or a magic spell's incantation—functional phrases whose value is in their effect, not their standalone literary merit. Copyright law has consistently denied protection to such short, functional strings of text.

A prompt is an instruction manual reduced to a single command. It tells the model what to do, not what to be. This functional nature places it outside the realm of copyrightable expression, just as we cannot copyright:

• "Turn left at the intersection"
• "SELECT * FROM users WHERE age > 21"
• "Preheat oven to 350°F"
• "sudo apt-get update"

These are all potentially valuable, precise, and skill-dependent utterances. None are works of authorship.

The Gas Station Analogy

Directions vs. Destinations

Consider these directions: "Take I-95 North to Exit 42, turn left at the light, the Shell station is 200 yards on your right."

These directions have several properties:

1. They describe a path through public infrastructure
2. They reference shared landmarks and conventions
3. They lead to a destination that existed before the directions
4. They can be expressed in infinite variations with the same result
5. Multiple people can independently derive similar directions
6. They are not copyrightable as works of authorship

Now consider this prompt: "Explain quantum entanglement using an analogy to dance partners, suitable for a high school audience."

This prompt has precisely the same properties:

1. It describes a path through semantic infrastructure (language, concepts, metaphors)
2. It references shared knowledge and conventions
3. It leads to conceptual territory that existed before the prompt
4. It can be rephrased in countless ways with similar results
5. Multiple people can independently derive similar prompts
6. It is not copyrightable as a work of authorship

Just as giving directions to a gas station grants you no ownership over the station, the roads, or the city's layout, crafting a prompt grants you no ownership over the conceptual territory, the language used to navigate it, or the model's capacity to explore it. More to the point: you cannot file a copyright claim on directions to a gas station because directions are functional, brief instructions, not expressive works. The same is true of prompts.

The Specificity Fallacy

One might object: "But my directions are very specific, very clever. I found a route that avoids traffic, uses back roads no one knows about."

This is the specificity fallacy—the belief that clever execution transforms description into property. But specificity changes nothing fundamental:

• Specific directions still traverse public roads
• Clever routes still use shared infrastructure
• Unusual paths still lead to places that already exist
• And clever directions remain functional instructions, not copyrightable expression

Similarly, a highly specific, cleverly crafted prompt:

• Still uses public language
• Still navigates shared semantic space
• Still points to conceptual regions that already exist in the model
• Still remains a functional instruction, not a work of authorship

The quality of navigation does not grant ownership of the territory, nor does it transform uncopyrightable material into copyrightable expression.

Multiple Discovery and the Illusion of Uniqueness

Convergent Navigation

In the history of exploration, multiple navigators often independently discovered the same islands, the same trade routes, the same passages. This is not coincidence—it is the natural result of navigating the same physical space with similar tools and goals.

The same is true in prompt space. Given:

• A shared language
• A shared model architecture
• Shared goals (e.g., "get the model to write better code," "make it explain complex topics simply")

Multiple prompters will independently converge on similar prompts. This has been empirically observed: the same prompt engineering techniques are "discovered" repeatedly by different users.

This alone should disqualify ownership claims. You cannot own what others would inevitably discover independently by navigating the same shared space.

The Banality of Most Prompts and Scènes à Faire

Furthermore, most effective prompts are variations on common communicative patterns:

• "Explain X as if I'm Y" (a standard pedagogical frame)
• "Write in the style of X" (a standard creative constraint)
• "Act as X and respond to Y" (a standard role-play structure)
• "List the top N examples of X" (a standard organizational request)
• "Summarize X in Y sentences" (a standard compression instruction)

These are not novel inventions. They are standard ways humans have been structuring requests and framing tasks for as long as language has existed. Applying them to AI systems does not magically transform them into intellectual property.

Legally, this is the scènes à faire doctrine at work. Standard, commonplace elements naturally associated with a topic or genre are not protectable. Most prompt structures are the scènes à faire of human-to-AI interaction—they are the necessary, expected building blocks of instructing a language model. You cannot own "Act as a..." any more than you can own "Once upon a time..." in fairy tales.

The Practical and Legal Impossibility of Prompt Ownership

The Substitutability Problem and the Merger Doctrine

Nearly any prompt can be rephrased in infinite ways while achieving functionally identical results:

Original: "Summarize this article in three bullet points, focusing on the main findings."

Variations:

• "Give me three key takeaways from this article."
• "What are the three most important points in this text?"
• "Extract the main findings as a brief bulleted list."
• "Condense this to its three core insights."
• "Bullet-point the top three discoveries here."

If someone claimed ownership of the first version, would all variations infringe? If not, the ownership claim is meaningless. If so, they're claiming ownership of the entire conceptual operation of "summarize into three points," which is absurd.

This is not just a practical problem—it's a doctrinal one. The Merger Doctrine in copyright law states that when there are only a limited number of ways to express an idea, the expression merges with the idea and neither is protectable. For many prompting tasks, there are only so many efficient ways to instruct the model. Allowing ownership of one version would monopolize the underlying functional goal itself.

When expression merges with function—when the way you say something is inseparable from what you're trying to accomplish—copyright law wisely steps back. Prompts live squarely in this space. The instruction "summarize this in three points" is not merely one expression among infinite possibilities; it is one of a small set of efficient expressions of a functional goal. Protecting it would mean protecting the goal itself.

Enforcement Would Require Owning Language

To enforce ownership of a prompt, one would need to claim ownership over:

1. The specific words used (but language is a commons)
2. The semantic content (but concepts are shared)
3. The structure of the request (but communicative patterns are universal and fall under scènes à faire)
4. The region of conceptual space accessed (but that space is collectively created)
5. The functional instruction itself (but functionality is explicitly excluded from copyright)

Each of these is impossible to own without asserting ownership over fundamental shared resources. It would be like trying to own the concept of "north" because you frequently give directions that involve traveling north.

Copyright protects expression, not ideas, systems, methods, or functional instructions. A prompt that achieves a particular result with an AI is fundamentally a method—a technique for producing an output. Methods are the domain of patent law (which requires novelty, non-obviousness, and utility), not copyright. And no prompt would survive patent scrutiny given the prior art of human linguistic interaction.

Prompts as Uncopyrightable Utterances

The Category of Functional Speech

Legally, prompts reside in a category of utterances that have never qualified for copyright protection. Consider these examples:

Commands:

• sudo apt-get update
• DELETE FROM users WHERE inactive = true
• "Alexa, set a timer for 10 minutes"
• "Claude, explain general relativity"

Queries:

• SELECT * FROM users WHERE age > 21;
• "What is the capital of France?"
• "Show me Italian restaurants nearby"
• "How do I fix a leaky faucet?"

Formulas:

• E = mc²
• F = ma
• a² + b² = c²
• "Total cost equals unit price times quantity"

Instructions:

• "Assemble Part A to Part B, then insert Tab C into Slot D"
• "Preheat oven to 350°F, bake for 45 minutes"
• "Turn left at Main Street, proceed for 2 miles"
• "Explain this concept using simple analogies"

Directions:

• "Head north on Main for 2 miles, then turn right"
• "Take the second exit at the roundabout"
• "Navigate semantic space toward poetic expression with Shakespearean constraints"

All of these are potentially valuable. All require skill to formulate effectively. All can be more or less elegant, efficient, or clever. None are "works of authorship" in the copyright sense.

A prompt like "Write a blog post about quantum computing in the voice of a 1920s gangster" is a complex, clever command. It demonstrates linguistic skill and creative insight into how to frame a request. But it remains a command. The skill is in engineering the command, not in authoring a protectable text.

The law has wisely never tried to own the imperative mood. We do not copyright:

• Questions ("What time is it?")
• Requests ("Please pass the salt")
• Orders ("Stop at the red light")
• Instructions ("Enter your password")
• Invocations ("Siri, call Mom")

These are functional uses of language. They accomplish tasks. They are not literary works meant to be appreciated for their expressive content. Neither are prompts.

Brevity and the Threshold of Originality

Even setting aside functionality, most prompts fail the basic threshold for copyright protection due to their brevity. Copyright law does not protect short phrases, slogans, or simple instructions. There is no bright-line rule, but courts have consistently held that protection requires a minimal level of creativity and expression that short, utilitarian texts simply cannot meet.

Consider what gets rejected by copyright:

• Advertising slogans ("Just Do It")
• Book titles ("The Great Gatsby")
• Product names ("iPhone")
• Simple instructions ("Push to open")
• Short phrases ("All rights reserved")

Prompts typically fall into this category. A prompt like "Summarize this article" is six words, purely functional, and contains no creative expression worthy of protection. Even longer prompts like "Write a comprehensive analysis of the themes in Moby Dick, focusing on the symbolism of the whale and Ahab's obsession, suitable for an undergraduate literature course" are just detailed instructions—complex commands, not literary works.

The Constitution grants Congress the power to protect "writings" by "authors"—terms that imply creative, expressive content. A prompt is an utterance, an instruction, a coordinate. It is not a writing in the constitutional sense.

The Creative Contribution Question

What Would Ownership Protect?

Intellectual property is meant to protect creative contributions. So we must ask: what is the creative contribution in a prompt?

Not the language: Natural language is prior art, collectively developed over millennia.

Not the concepts: The ideas being discussed already exist in human knowledge.

Not the structure: Imperative sentences, question formats, and instruction patterns are universal linguistic tools.

Not the model's response: The response emerges from the model's training, not from the prompt alone. The prompt triggers but does not create the output.

Not the technique: Most prompt engineering techniques are either obvious applications of communicative norms or are independently rediscovered. The scènes à faire doctrine excludes these from protection.

Not the function: Copyright explicitly does not protect methods, systems, or processes—only expression. A prompt is fundamentally a method for eliciting a response.

What remains? At best, a particular arrangement of common elements to achieve a common goal. This is so minimal that it falls below any reasonable threshold for intellectual property protection. It's like claiming ownership of the sentence "Please pass the salt" because you arranged common words in a common pattern to achieve a common goal.

The Difference Between Craft and Property

There is a crucial difference between:

1. "I am skilled at crafting prompts" (a claim about craft)
2. "I own this prompt" (a claim about property)

The first is legitimate. Prompt engineering is a skill. Some people are better at it. They understand the model's patterns better, phrase requests more effectively, achieve better results more consistently. They can and should be compensated for their expertise, their time, their effective techniques.

But skill does not imply ownership. A skilled navigator does not own the routes they travel. A skilled chef does not own the recipe structure "combine ingredients, apply heat, serve." A skilled prompter does not own the coordinates they've discovered in semantic space.

Copyright protects the fixed expression of creative works, not the skill involved in producing functional utterances. We can recognize and reward prompt engineering expertise without granting monopolies over language itself.

Collective Creation and the Commons

Who "Owns" the Conceptual Space?

If anyone could be said to own the conceptual space that prompts navigate, it would be:

• The millions of authors whose work trained the model
• The speakers of the language who shaped its semantics
• The researchers who developed the architecture
• The annotators who refined the model's behavior
• The companies and institutions who funded the work
• Humanity collectively, as the source of language and knowledge

But even this is misleading, because ownership implies exclusion, and this space is non-rivalrous. My use of a conceptual region does not diminish yours. My prompt does not consume the semantic space it navigates.

The proper framework is not ownership but contribution to a commons. Every human who has used language has contributed to the space that language models encode. The space belongs to all of us, which is to say it belongs to none of us exclusively.

Prompts as Annotations on Shared Infrastructure

The closest analogy might be to annotations in a library book. If I write helpful notes in the margin of a library book, I have contributed something useful. Others might benefit from my notes. But I do not thereby own:

• The book
• The page
• The concept discussed
• The practice of annotation
• The library itself

I have made a contribution to a shared resource. That contribution might be valuable. It demonstrates skill and insight. But it remains within the commons. It does not transform into property merely because it is useful.

Similarly, a well-crafted prompt is a useful annotation on the shared infrastructure of language and AI. It points to productive regions of semantic space. It demonstrates skill in navigation. But it cannot be owned any more than marginalia can be owned, because it operates entirely within and upon a collectively-created commons.

Conclusion: From Philosophy to Law

Claims of prompt ownership fail on multiple grounds:

Philosophically, they mistake:

• The map for the territory
• The coordinates for the destination
• The description for the thing described
• Navigation for creation
• Skill for property
• Contribution for ownership

Legally, they fail because prompts:

• Are functional instructions, not expressive works
• Fall below the threshold of originality due to brevity and utility
• Consist primarily of scènes à faire—standard patterns of human-AI interaction
• Are subject to the merger doctrine where expression and function become inseparable
• Belong to a category of utterances (commands, queries, instructions) that has never received copyright protection
• Cannot meet the constitutional definition of "writings" by "authors"

The proper framework recognizes prompts as:

1. Functional instructions for navigating semantic space
2. Expressions of navigational skill, not authorship
3. Coordinates in collectively-created conceptual territory
4. Tools for accessing what already exists
5. Categorically uncopyrightable utterances under established legal doctrine

We can value prompt engineering without claiming ownership. We can compensate prompt crafters for their expertise without granting property rights over language itself. We can build on each other's discoveries without permission because we are all navigating the same shared space that humanity created together—and because the law has wisely never allowed functional instructions to be monopolized through copyright.

To claim ownership of a prompt is to claim ownership of the gas station because you gave someone directions. The station was already there. The roads were already there. The city was already there. You just knew how to point the way.

More than that: your directions themselves are not copyrightable. They are functional speech—part of the free fabric of human communication, the imperative and instructive discourse we use to operate the world and now, the minds we've built.

That fabric must remain free. Not just because it would be philosophically absurd to claim ownership of coordinates in collective space, but because the law has always recognized that functional instructions, brief commands, and navigational utterances belong to everyone.

AI prompts are uncopyrightable functional instructions for navigating collectively-created semantic space. This is not merely philosophical preference—it is settled legal doctrine. The gas station directions aren't just philosophically un-ownable; they're also not something you could ever file a copyright claim on in the first place.

---

"The map is not the territory, the direction to the treasure is not the treasure itself, and the imperative mood has never been property."