How to Copyright Your AI Creations

Published on
Product Minting

AI content isn't naturally copyrightable, but there's a workaround

As a content creator, you might be tempted by generative AI's sexy promise: slashed costs, productivity on steroids, and creativity beyond imagination.

Hold up. Don't go all in just yet. Every temptation comes with a cost. And it's not just the subscription fee you pay.

It's the ownership over your AI creation that you're unlikely to secure.

The thing is…

What you create with AI is generally considered non-copyrightable.

That’s not a big deal if you don’t care about cashing in on your content.

Feel free to make the machine your secret slave and score big on conveyor productivity.

But if you do, think twice


The hours you pour into crafting prompts and wrestling with randomness could end up as public domain material—free for anyone to swipe and repurpose.

Wouldn’t it be a bummer?

However, if you use AI wisely, you can dodge that bullet.

Next up, let’s explore why AI and copyright clash and how to use the emerging tech without sacrificing ownership.

Note: While I'm diving deep into the topic, don't mistake my rant for legal advice. I'm here to inform and provoke thought, that's it.

A few core reasons here.

You’re not the Puppet Master of AI

You might think you're in control when you feed AI with detailed prompts, but the reality is a bit more complex.

In fact, how AI interprets and executes your instructions is largely out of your hands. Not to mention the dataset it uses.

Between your prompt and the final output, there are numerous variables at play:

encoder and decoder layers, attention mechanisms, hyperparameters, and a dash of randomness—all elements you're not directly controlling.

General representation of the Transformer architecture in an LLM. Image credited to Vitaflux

General representation of the Transformer architecture in an LLM. Image credited to Vitaflux

Even experts admit that the inner workings of LLMs (Large Language Models) are a ‘black box’.

So, you can be super precise with your prompts, but it's AI that's calling the shots on the output, not you.

That's what makes claiming full authorship of AI content so tricky.

This stance is pretty much the global consensus among IP watchdogs. If you want to back this up and bore yourself with legalese, for example, check out the US or EU doctrines on the IP&AI tandem.

Now onto the second big hurdle.

Predicting the fallout of opening this Pandora's Box is a real head-scratcher right now.

The surge in lawsuits over 3rd-party rights infringements by AI tools, coupled with the ongoing ethical debates around emerging tech, has regulators pumping the brakes on embracing innovation.

You've probably heard about artists suing Stability AI, Midjourney, Copilot, Meta, and OpenAI.


These projects used artists' works to train their datasets without permission and are now cashing in on it.

Artists are upset.

The outcomes of these lawsuits are still up in the air.

Even if they set precedents, it's unlikely to clear up the AI copyright fog entirely.

Given that the AI boom is still in its infancy, it's tough to see what lies ahead. If regulators jump the gun, who knows what kind of mess we could end up in?

They're better off dodging this headache and waiting for the landscape to somewhat stabilize.

A Glimmer of Hope

However, regulators don't want to look like Luddites holding back progress.

So, they're playing it safe with a vague stance:

'We'll decide on a case-by-case basis.

Prove that your work is AI-assisted, not AI-generated.'

Sounds, like the ball's in your court.

How exactly to prove it is a million-dollar question - of course, there’s no ‘how-to’ guide on this so far.

On the other hand, if something doesn't exist yet, we can try to invent it ourselves.

For that, first, let's dive into this AI-assisted vs. AI-generated dilemma.

AI-Assisted vs AI-Generated: What's the Difference?

The dividing line here is the amount of human effort.

As you might guess, in AI-generated work it’s scant, while in AI-assisted work, it takes the lead.

There's no clear-cut threshold to quantify this effort, which muddies the waters even more. Also, remember the premise that a creator has little control over AI?

So, how do you make sense of this murky 'AI-assisted' category and apply it to your work?

Here's a thought: use the ‘AI-assistance test’ (that's my term for it, for simplicity) from the EU IP doctrine.

In my opinion, it's the clearest framework out there to cut through the fog.

Heads up: While this test might not apply directly to non-EU folks, it can still guide you on how to structure your AI work to fit the "AI-assisted" label.

Still better than flying blind, IMO

Next is what the EU Test actually suggests—no legalese, I promise.

The AI-assistance Test

Image created by the author

Image created by the author

The Test lays out a few criteria to measure your work against, ensuring it can be classified as AI-assisted:

  1. Literary, scientific, or artistic domain
  2. Human effort
  3. Originality & creativity (the crucial one)
  4. Expression

The first one is pretty self-explanatory, so let's unpack the other three.

1) Human Effort

As mentioned, there's no fixed ratio or percentage that defines human effort as dominant in the creative process. It just needs to be the driving force, as they say it.

The irony of this criterion is that we're not living in a sci-fi world yet where robots handle all the creative work from start to finish.

The current reality is that human effort is, in most cases, quite significant.

How much is enough? That's the big question.

Perhaps, the next criterion could help us crack this puzzle.

2) Originality and creativity

The Tesr says a work is original and creative if it shows the author's personality and creative choices at each AI-assisted stage.

Keep "creative choices" in mind - that's exactly what you'll need to meticulously document along the way.

Specifically, you'll want to explain your choices at every creation stage:

  • Conception
  • Execution, and
  • Redaction

Let’s dig a bit deeper into each.

a) Conception

At this stage, you've got to show that the spark of the original idea came from your own creative mind, not the AI.

For example, your creative choices could look like:

  • The way you approach your research
  • How you select material, be it subject matter, plot, or even melodic ideas
  • Your unique spin on genre, style, format, and technique
  • How you tie this work into your previous ones to maintain a consistent vision
  • Leveraging your personal story or expertise for a unique angle

Maybe you used your own non-AI visuals as prompts, or perhaps you've got interviews that can vouch for your creative input.

So, before you sync with AI, think about how you'll detail these kinds of creative choices that make your concept truly yours.

b) Execution Phase

This stage is primarily in the AI's court focusing on how it interprets and executes your prompt.

However, you're not entirely hands-off.

Here your creative choices can revolve around monitoring the output, adjusting AI settings to align with your tone, and instructing it to stick to your original concept.

Document them in detail.

The aim here is to show that you steered the AI to follow your original vision, rather than letting it call the shots.

c) Redaction Phase

Now, let's talk final touches.

This is where you make sure the AI didn't go rogue and your end product does reflect your original concept.

Depending on what you're working on—be it an article, a video, or a design piece—you might be doing a whole bunch of things here: rewriting, editing, formatting, color-correcting, etc.

Basically, all the nitty-gritty stuff that happens before you hit "publish" and start marketing it.

Make it clear how you polished your work up to its final form.

3) Expression

The last piece of the copyright puzzle is making sure your creativity and personality are indeed expressed in the final product.

You want to make it clear that the outcome isn't some AI improvisation and your original intent didn't get lost.

All your authorial creative choices should shine through your end work.

Now let’s wrap it all up

If you're aiming for your work to be copyrightable, you've got to prove it's AI-assisted.

Approach your work through the EU Test lens we’ve just gone through in this article and document your creative choices at every step.

Your backup documentation should essentially say:

'Hey, I'm the creative genius here, and AI was just my sidekick.

This is how my work embodies my tone, style, and vision—things I've been doing long before AI came into the picture (the last part is ideal).

I made my creative choices at every stage—conception, execution, and redaction—and they're loud and clear in the final product.'

Sure, this isn't a guarantee your work will be stamped as yours.

There are other factors at play, not just the AI part.

But knowing this framework gives you the best shot.

From there, you'll see how it goes.

Good luck!

*If you enjoyed this article, you'll likely vibe with me on Twitter. Hit me up there and let’s chat!

Discussion (20)

Not yet any reply