Imagine this scene: It is 2:00 AM on a Tuesday. A startup founder, let’s call her Elena, is hunched over her laptop, coffee going cold on her desk. Over the past three months, she and a lean team of three have built what would have taken a thirty-person engineering and design firm a full year.
With Claude refactoring complex algorithms, GitHub Copilot generating clean Python architecture, and Midjourney generating cinematic brand illustrations, they built a breakthrough SaaS platform. They launch. Users flood in. Within six months, a venture capital term sheet lands in Elena’s inbox valuing the company at twelve million dollars.
Elena feels triumphant. She has harnessed the single greatest productivity leap of the century.
Then comes week three of legal due diligence.
The lead investor’s IP counsel sends an eight-word inquiry: “Please provide chain of title for core assets.”
Elena’s attorney begins looking into how the software was assembled, where the artwork originated, and how the proprietary visual branding was crafted. And then comes the phone call every founder dreads:
“Elena, you have a brilliant business. But according to federal law, you don’t actually own the copyright to your code or your visuals. Anyone in the world can copy them tomorrow, and you cannot stop them.”
The silence on the line is deafening.
The Great Creative Misunderstanding
We are living through a massive collision between technological capability and legal reality.
Every single day, brilliant creators, startup executives, and enterprise teams feed prompts into AI models, watch polished text, imagery, and code materialize in seconds, and naturally assume: “I prompted this into existence. It is mine.”
It feels intuitive. After all, you spent forty-five minutes tweaking descriptive tokens, setting parameters, adjusting temperature, and rerolling seeds until the machine produced exactly what you envisioned. You invested sweat, intellect, and vision.
Why shouldn't it be yours?
Because under the law, the copyright system does not care how much sweat went into the prompt. It cares about who held the creative brush.
The Bedrock Rule: Machines Do Not Have Creative Souls
To understand why Elena’s term sheet hit a wall, we have to look back more than a century before the first GPU was ever fired up.
In 1879, the Supreme Court ruled in the landmark Trade-Mark Cases that copyright protection extends only to intellectual creations that are “founded in the creative powers of the mind.”
For 145 years, through the dawn of photography, player pianos, synthesizers, and CGI software, federal courts have held onto one immovable anchor: copyright requires a human being.
If a monkey snatches a photographer’s camera and snaps an accidental, world-famous grinning portrait (the infamous Naruto v. Slater “monkey selfie” litigation), that photo belongs to everyone—and therefore to no one. It is in the public domain.
When computer scientist Stephen Thaler spent years suing the U.S. Copyright Office to register visual artwork generated autonomously by his AI system, the federal courts shut the door with absolute clarity in Thaler v. Perlmutter. Judge Beryl Howell reaffirmed that human authorship is an indispensable, non-negotiable prerequisite of American copyright.
In March 2023, the U.S. Copyright Office laid the issue to rest in formal guidance: When you give a prompt to an AI model, the model—not you—is making the expressive choices.
Why "Prompting" Isn't Authorship in the Eyes of the Law
Here is the easiest way to understand how federal copyright examiners view prompt engineering:
Picture yourself walking into the studio of a world-class oil painter. You hand her a three-page creative brief:
“Paint a weathered lighthouse perched on a jagged cliff at dusk. Use the rich color palette of J.M.W. Turner. Make the waves crash dramatically against granite rocks, and paint a single beam of golden light piercing through violet storm clouds.”
The artist spends three weeks at the easel, blending pigments and applying brushstrokes, and delivers a masterpiece that matches your brief to perfection.
Who owns the copyright in that painting?
Under standard copyright law, the artist does—not you. You provided instructions, vision, and critique. But the painter exercised the human, expressive control over the canvas.
When you type a prompt into an LLM or an image generator, the federal government views you as the client commissioning work, and the algorithm as the commissioned painter. The machine executed the expressive choices. But because the machine is not a human, no copyright is born at all. The work drops straight into the public commons.
The Zarya Tale: The Blueprint for Modern Creative Teams
Does this mean all AI-assisted work is doomed to sit unprotected in the public domain?
Not at all. The answer lies in how humans and algorithms divide the labor.
Consider the story of Kristina Kashtanova and her graphic novel, Zarya of the Dawn. Kashtanova wrote a compelling narrative and used Midjourney to generate stunning comic panels matching the storyline.
When the Copyright Office reviewed the registration, they delivered a groundbreaking decision that serves as the blueprint for every business today:
The Human Story & Script: Protected by copyright. A human wrote the words.
The Layout, Pacing, and Arrangement: Protected by copyright. A human curated, edited, and orchestrated which panel appeared where.
The Raw Midjourney Images: Registration canceled. Because they were prompted outputs, anyone can legally grab those individual image files and reuse them without infringement.
This distinction is where the entire game is won or lost.
The Hidden Landmines for Modern Businesses
For an individual hobbyist, whether an AI image is copyrightable is an academic curiosity. For a business, it is an existential balance sheet issue.
1. The Due Diligence Nightmare
When venture capitalists, banks, or corporate acquirers perform IP audits during financing or M&A, they require warranties confirming that your company has unassailable, exclusive title to its core assets. If your core software modules or digital product designs were generated wholesale by unguided AI, your exclusivity evaporates.
2. The Copycat Immunity
Imagine spending hundreds of thousands of dollars branding a DTC product line using photorealistic AI art. Six weeks after your launch, a drop-shipper copies your exact marketing renders, prints them on their own packaging, and undercuts your pricing by 40%. When you call an attorney to send a cease-and-desist, you discover you cannot enforce an unregistered or uncopyrightable public-domain work.
3. The Registration Trap
You cannot file a federal copyright lawsuit in the United States without a registration certificate in hand (Fourth Estate Public Benefit Corp. v. Wall-Street.com). If you conceal your use of AI on Form CO, your registration can be invalidated in federal court for fraud or misrepresentation.
Building a "Human-First" Moat: How to Protect Your Work
So how do you harness the speed of AI without sacrificing ownership of your intellectual property? You must consciously design your creative and development pipelines around human creative control.
Here are three rules every founder and creative director should implement today:
Rule 1: Treat AI as the Rough Draft, Never the Final Product
Let AI generate the scaffolding, the boilerplate, or the initial brainstorm. But ensure that your human engineers, writers, and designers substantively refactor, sculpt, and rewrite the final expression. Copyright attaches to original human modifications. The more manual, creative delta you introduce between the raw prompt output and the final asset, the stronger your legal shield.
Rule 2: Keep a Digital "Provenance Paper Trail"
If your IP is ever challenged, you need to prove where the machine ended and where the human began.
Maintain version-controlled Git commits that clearly document human refactoring.
Save iterative design project files (Figma, Photoshop, Illustrator) with editable human-crafted layers, typography, and brush adjustments.
Keep intermediate drafts that demonstrate human selection, curation, and structural transformation.
Rule 3: Be Relentlessly Honest with the Copyright Office
When submitting registrations for hybrid works on Form CO, transparency is your best defense. Use the Limitation of Claim section to disclaim raw AI components while explicitly claiming ownership over the human-written text, architectural compilation, and creative post-processing. A narrow, bulletproof copyright is infinitely more valuable than an overbroad registration that collapses under judicial scrutiny.
Final Thoughts: The True Moat Is Still Human
Generative AI gives modern teams superpowers. It eliminates blank-canvas paralysis and allows small, agile teams to compete with industry giants. But speed should never be confused with legal title.
The companies that thrive in this new landscape won’t be the ones who blindly press "generate" and assume they own the future. They will be the ones who pair cutting-edge tools with rigorous legal foresight—recognizing that in the eyes of the law, true value still begins and ends with the human mind.
Disclaimer: Practice in Florida limited to Federal Immigration and Federal Intellectual Property matters. Admitted only to the State Bar of Texas. Attorney advertising; prior results do not guarantee a similar outcome.