Creating With AI: Where Copyright Protection Stops or Thins

Every few weeks another artificial intelligence engine is introduced to the public that empowers creators to new heights of creativity. Tools like ChatGPT, Midjourney, Nana Banana Pro, and DALL-E are becoming mainstays of working artists’ daily workflow. Artificial intelligence is at once both empowering and disempowering. Boosts in productivity and expanded creativity are offset by ambiguities regarding intellectual property ownership and anxieties about future career stability. Creators across New York, Los Angeles, and Georgia are inundated with daily reminders of the actual “great replacement theory”, that AI outputs will eclipse human ones in creative spaces and beyond. According to recent industry surveys, the majority of creators are unsure whether and to what degree works they create with AI belong to them when publishing content online!

More so, many people believe the widespread misconception, “If I made it with AI, I own it” or that alternatively authoring AI prompts is sufficient to warrant protection. That belief is deeply flawed. Copyright law has long upheld the principle that protection requires human authorship – no “ifs, ands, or buts”. Animals, for instance, don’t enjoy copyright protection whether it’s elephants with paintbrushes or macaques taking selfies. Works created solely by a machine have historically been ineligible for copyright protection. Human use of AI is creating novel issues of law and what may be an evolution of longstanding copyright principles.

In this article, I’ll explain where copyright protection stops or thins when using AI in your creative process, the current state of AI & copyright law, and tips to creating with AI so that your work stays owned and protected.

What Copyright Law Protects (and What It Doesn’t)

U.S. copyright law protects “original works of authorship”. The law requires that the work be independently created by a human with a minimal degree of originality. The threshold for originality is surprisingly low and seemingly subjective. The mandate of human authorship, however, has been historically strict. To qualify for copyright protection, the human contribution must be substantial and rise above superficial. The human author must make creative choices that meaningfully shape the outcome.

Non-human or mechanical outputs that lack meaningful human involvement do not meet the threshold for copyright eligibility. To be real, anything involving “artificial” creation would seem unlikely to meet a human authorship criteria. Yet since all life and creation appears headed towards a relationship with artificial intelligence, it’s not as simple as delineating artificial from human creation anymore. Instead, copyright law is moving towards the necessary evolution of creating a legal framework for protection of works involving a combination of human and artificial intelligence. For the time being, those directives are coming from the U.S. Copyright Office while we wait for the U.S. Congress (bless its heart) to provide federal legislation.

The State of the Human Authorship Requirement

As of February 2026, if a work is generated entirely by AI, it’s still ineligible for copyright protection. Creative prompting still doesn’t constitute human authorship under the law. The Copyright Office’s rationale is that AI-models generate the work exclusively within their computation engines without human involvement. To be fair, there’s a viable argument that creative, attenuated prompting involves original authorship. The better or more creative your prompting, the better your AI outputs. Accepting prompts as original works of authorship would create a flood of issues at the Copyright Office. If prompts do become copyright-protectable, it’s likely something that the law will move towards slowly and qualify with a great degree of specificity. Rightly so, one precaution is preventing the corporate owners of AI engines, like Microsoft or Anthropic, from claiming broad copyright ownership over the prompts of its users.

How AI-Generated Content Is Treated Under Current U.S. Law

The U.S. Copyright Office has released three separate reports outlining its conclusions over the current state and future of copyright law and AI. Released over a year and a half (2024-2025), the reports cover the Office’s conclusions regarding digital replicas, the copyrightability of AI, and large generative AI training.

At the heart of its conclusions, the Copyright Office has stated (or reiterated) that:

Works must have a human author to be copyrightable
AI tools can assist the creative process, but cannot be the creator
Simply crafting a prompt isn’t enough to confer authorship.

This position aligns with federal appeals court rulings confirming that purely AI-generated art cannot be copyrighted because it has no human author. A recent D.C. Circuit Court of Appeals decision denied copyright protection to a strictly A.I.-generated work because as the presiding judge ruled, “only humans can be incentivized to create” and the AI-claim before him was effectively trying to put the “cart before the horse” by asking for protection from laws that don’t exist yet. Thaler v. Perlmutter, No. 23-5233 (D.C. Cir. 2025). The case was a litmus test for the status of AI copyrightability, which the court soundly rejected. The case is currently being appealed to the Supreme Court but has not yet been granted certiorari for SCOTUS to hear the case.

As of this writing you must disclose or disclaim the AI elements of your work when registering for copyright protection. This means you’ll need to acknowledge that non-ownership of the AI elements of your work, just the human ones.

How AI Creates the Following (And Your Legal Risk)

Understanding copyright’s human authorship requirement will help ensure that your AI-assisted creative works rise to the standards of copyright eligibility. It will allow you to substantiate the human element of your AI-assisted works when you record your works and apply for registration.

In real-world applications, it’s becoming increasingly difficult to screen for the human element of AI works. The onus is on the registrant to prove “meaningful” human contribution when registering a work. If you’re tempted to lie through the process, just remember you’d still have to prove this contribution if you ever have to defend against infringement.

Lyrics, Scripts, Dialogues, Articles

If you’re using AI to brainstorm or create drafts, those outputs are not independently protectable. But there are a few ways to use AI properly for copyright protectability, especially if you still care about the artistic contributions to your own works (which some success-seekers do not).

The best way is to take the AI-outputs and revise them in your own voice to an extent that you can directly articulate the changes you made. This could be changing just a few words, the order of the words, or using the output as inspiration and rewriting it in its entirety.

You could also plug AI outputs (individual lyrics or script dialogue) into your works and still be protected provided that the writing around was and remains substantially human. The extent you do this is up to you and arguably a creative direction in and of itself. Simply stated, the more you contribute or alter, the stronger your copyright protection will be.

Melodies, Arrangements, Videos, Pictures

The same principles apply to non-linguistic outputs – melodies, visuals, pictures, and their particularized arrangements. Melodies and arrangements are generally easier to work with since you can rearrange them with relative ease. Changing note placements and contour will constitute human authorship so long as you can show those contributions are significant.

Video or picture outputs work the same way but can be more difficult to alter or adjust. You can add meaningful contributions in post-production – like editing for style, color grading, or splicing – but you’d only be protected as to those changes and not the remaining content. This means your copyright protection would be thinner than if you created the work yourself and then used AI for post-production which is preferable.

If you’re working with AI as an assist rather than de facto creator, your copyright protection will be stronger. If using a platform like Suno for instance, it’s best if you contribute an articulable musical idea upfront (and recorded prior separately) that you build around rather than simply text inputting and presuming you can copyright the output thereafter. This would be the same with other audiovisual AI engines as well.

Best Practices

As a creator, it’s best if you devise a method that’s in a proper order of operation for you to show your provable human creative contribution. Write an idea partially or in full first and then use AI to assist in your creative outputs from there.

We’re headed towards an era where this may not matter, but if you ever need to protect your works, it’s more persuasive if you can show your works in a written, non-AI form and then articulate exactly how you used AI in your creative process. Legal protection or enforcement will ultimately require you to show the visible or provable human contributions to your work.

Where Copyright Protection Fails

Fully AI-Generated Outputs

If a work is created autonomously through AI or solely through a creative prompt without any significant human contribution thereafter, it’s ineligible for copyright protection. Multiple prompts towards the same work won’t strengthen your odds of eligibility.

Minimal Human Input

The key criterion for human contributions according to the U.S. Copyright Office is that they’re “substantial, rising above mere superficial”. This will be the working standard as we move into a world where AI works become indistinguishable from human ones. For now, it’s important to remember that you’ll need to articulate a “substantial” amount of human involvement for a successful registration and beyond, for successful infringement defense. Without this, you can still commercialize your sole AI or AI-assisted works, but there’s just nothing to prevent another from monetizing them to your loss.

Ownership Enforcement Problems

As I just mentioned, without legal protection, you’d lose rights in the monetization of your works. You also lose other exclusive rights under copyright law like the right to reproduce, distribute, or license your works. If you’re unable to show the distinct human contributions to your work, it will create a serious enforcement issue. This will become especially relevant if the law ever moves towards a standard where the presumption is that all works are AI-created and the burden of proof is on the creator to prove otherwise. It’s clear that some measures will need to be taken to heighten standards for protection especially if distinguishing AI works becomes impossible thought there’s no certainty when. There may come a time where we live in a world where copyright protection doesn’t exist, an extreme but increasingly plausible outcome.

Practical Tips for the Time Being

Persuasively Document

Document your personal contributions throughout your processes – from initial drafts, editing, arrangement choices. You could do this in a number of ways, but stick to the most compelling ones backwards – a dated video or voice note of you creating your work can be more persuasive than in written form. Make sure you’re using something that is time-stamped. You should also go on record with your personal AI methodology especially if you can support it with actual evidence of process. If it’s well understood how you use AI in your creative process and you have direct evidence to substantiate it, then you’ll be able to prove ownership much easier.

Licenses and Commissions

For licensing or commissioning works, having the assurances I mention in the paragraph above can be tremendously helpful. You’ll also want to provide contingencies in your contractual language to ensure that the work you’re licensing has a distinguishable human author and that any works commissioned are provably eligible for copyright protection. It may not matter as to the work product itself if a license relies too heavily on AI, but it will matter for enforcement, monetization, and your control of the work. It will also assuage your licensees who will want assurances that the work is under your lawful ownership.

The AI Age is Here

AI is a tool with otherworldly creative potential, but it also threatens to devalue works and displace entire creative fields. The U.S. Copyright Office is slowly crystallizing how copyright eligibility will work in the age of AI, but there’s still tremendous uncertainty about the implications for copyright and intellectual property law as a whole. For the time being, the law is still holding to long-standing principles about human authorship (as it should), but it’s not hard to see how the AI boom will bring into question whether the Copyright Office can effectively navigate the human/AI conundrum regarding creative works. We can all imagine the de-privatization of entire creative fields as AI begins to predominate autonomously or the Copyright Office determines that distinguishing human elements is just not possible.

For the time being, you should take the information you learned in this post and consider what your significant human contributions will look like in the context of your AI-assisted works. A common suggestion I offer is to be able to articulate with some clarity (to an attorney for instance) the human elements of your creative process and then think over how you could prove that with direct or circumstantial evidence.

In the meantime, we’re all collectively waiting for new developments from the U.S. Copyright Office or federal government on how copyright law will adjust to the age of AI. When changes do happen, you can always revisit our blog for new updates or reach out to our firm for advice.

At www.zalaiplaw.com, we help creators and rights holders in Georgia, New York, and California navigate copyright risks, draft enforceable agreements, and protect their intellectual property in the era of AI.