whoownsthecode()
resources.md ~/whoownsthecode

Resources

Copyright is a legal protection automatically granted to the original creators of software code, giving them exclusive rights to use, modify, distribute, and license their work. For developers and companies, owning the copyright ensures control over how the software is used and monetized. It helps prevent unauthorized copying or stealing of the codebase, making it a critical asset for building valuable, defensible products and attracting investors. Without proper copyright ownership, your business could face legal risks or lose its competitive edge.

Source Code as a Literary Work

Under copyright law, source code is treated as a literary work, just like a novel or screenplay, because it’s a written expression of ideas in a specific, fixed form. This classification means that the structure, organization, and actual lines of code are protected from unauthorized copying. Recognizing source code as a literary work reinforces its value as intellectual property and strengthens legal claims in the event of infringement.

By default, the person who writes the code owns the copyright. For W2 employees, the employer usually owns the code under “work made for hire.” In contract work, ownership depends on the agreement. Unless it clearly assigns rights, the developer may keep them. Clear contracts are key to avoiding confusion.

A copyright exists the moment a human fixes original code in a tangible medium, but in the United States you cannot sue for infringement until the Copyright Office has issued a registration. The Supreme Court confirmed in Fourth Estate v. Wall-Street.com (2019) that a pending application is not enough. Timing also matters: statutory damages and attorney’s fees are available only if the work was registered before the infringement began or within three months of first publication. Registration costs about $65 per work, and the application asks who the author is, which is where AI-generated code runs into trouble. A © notice is optional and is not a substitute for registration.

Open Source != “No Copyright”

Open source software is still protected by copyright. The difference is that the author grants permission to use, modify, and share it under specific license terms. Copyright ownership remains with the creator or project maintainers. Using open source doesn’t mean the code is free of legal restrictions; it just comes with a different set of rights and obligations.

The U.S. Copyright Office’s guidance and reports on AI and copyright reaffirm that copyright protection requires human authorship. Its 2023 registration guidance, followed by its 2025 report on the copyrightability of AI outputs, clarified that while AI-assisted works may qualify, only the portions with meaningful human creative input are eligible. These serve as guidance for creators and companies navigating copyright in the era of generative AI, especially in fields like software development, where human-AI collaboration is increasingly common.

For a chronological view of these and other events, see Developments.

Naruto v. Slater (the “monkey selfie” case)

Long before generative AI, the courts addressed whether a non-human can own a copyright. In Naruto v. Slater, PETA sued on behalf of a crested macaque that had taken a selfie with a wildlife photographer’s camera, asking the court to declare the monkey the copyright owner. The Ninth Circuit held in 2018 that animals have no standing under the Copyright Act, and the Copyright Office’s own rules exclude works produced by animals or machines without human authorship. The principle is the same one now applied to AI-generated code: the author must be a human, no matter who owns the laptop, the repository, or the AI subscription that produced the code.

Thaler v. Perlmutter

In Thaler v. Perlmutter, the D.C. Circuit ruled that works created entirely by AI aren’t eligible for copyright and that only humans can hold copyright protection. Thaler tried to register AI-generated art, but was denied, reinforcing that meaningful human authorship is required. This principle also applies to software code: AI-generated code without human creative input may not be protected by copyright.

Supreme Court Declines to Hear Thaler

On March 2, 2026, the U.S. Supreme Court denied certiorari in Thaler v. Perlmutter, declining to review the D.C. Circuit’s ruling. A cert denial isn’t an endorsement of the lower court’s reasoning, but the practical effect is significant: the human authorship requirement now stands as settled U.S. copyright law unless Congress acts. For software teams, this removes any near-term possibility that purely AI-generated code gains copyright protection on its own. Meaningful human creative input remains the price of ownership.

Allen v. Perlmutter

The open question is how much human input is enough, and that line applies directly to heavily prompted code. In Allen v. Perlmutter, an artist who used more than 600 iterative prompts to produce an image was denied registration by the Copyright Office and is now challenging that refusal in the U.S. District Court for the District of Colorado.

Zarya of the Dawn

The Copyright Office has already shown how it treats mixed human and AI works. In February 2023 the Office decided the registration for Kris Kashtanova’s comic Zarya of the Dawn, made with Midjourney images. It registered the human-written text and the human selection and arrangement of the images, but excluded the individual AI-generated images from protection. This is the template for mixed works: protection attaches to the human-authored layers, not the AI output. For software, the parallel is a mixed codebase, where human-written portions and the human selection and arrangement of components may be protectable while purely generated functions are not.

Thaler v. Vidal

Patent law reached the same answer as copyright. In Thaler v. Vidal, the Federal Circuit held in 2022 that an “inventor” under the Patent Act must be a natural person, so an AI system cannot be named as the inventor of a patent, and the Supreme Court declined to review the decision in 2023. USPTO guidance permits patents on AI-assisted inventions where a natural person contributed to the conception of the invention; an invention with no human inventive contribution is not patentable. For companies hoping patents might protect what copyright cannot, the result is the same: a purely AI-conceived invention is not patentable, and naming a human who did not actually invent it risks invalidating the patent later.

Thomson Reuters v. Ross

In Thomson Reuters v. Ross, the legal publisher sued Ross Intelligence for allegedly scraping and using its copyrighted legal content to train an AI legal research tool. Ross argued that its use was “intermediate” and transformative, not a direct republishing of Westlaw materials. The case highlights the legal tension between data scraping and fair use in AI training, raising important questions about whether using copyrighted text to train models constitutes infringement or innovation. In February 2025, the district court rejected Ross’s fair use defense and found infringement, the first decision to deny fair use for AI training. The Third Circuit heard oral argument on the appeal (No. 25-2153) in June 2026, and its ruling will be the first federal appellate word on the question. The district court expressly limited its ruling to non-generative AI, so its reach into generative-model training remains unresolved.

Doe v. GitHub

The only major U.S. litigation specifically about AI coding tools and their code output. A class action was filed in November 2022 in the Northern District of California against GitHub, Microsoft, and OpenAI over GitHub Copilot. The core allegation is that Copilot reproduces licensed open-source code without the attribution, copyright notices, or license terms the original licenses require. On September 16, 2026, the Ninth Circuit (No. 24-7700) affirmed dismissal of the DMCA claims: the tools generate new works that never contained the plaintiffs’ copyright management information, and stretching the DMCA to cover substantially similar output would supplant ordinary copyright law and its damages rules. The court took no view on whether substantially similar output supports an ordinary infringement claim, so that question is open. The breach-of-contract claims over the open-source licenses remain pending in the district court and are now the live core of the case. The court also rejected the district court’s requirement of an identical copy: minor cosmetic changes do not protect a defendant who substantially reproduces a work and strips its copyright management information, because the line is generation versus copying, not identical versus modified. And the panel decided only the output theory. Whether removing copyright management information from code before using it as training data violates the DMCA was not preserved below and remains open (Doe v. GitHub, Inc., No. 24-7700, slip op. at 13-17 (9th Cir. Sept. 16, 2026)).

Trade Secrets Don’t Require an Author

Copyright requires a human author (Thaler v. Perlmutter). A patent inventor must be a natural person (Thaler v. Vidal). Trade secret law has neither requirement. The Defend Trade Secrets Act, signed May 11, 2016, created a federal civil claim for trade secret misappropriation, and protection turns on secrecy, reasonable measures to keep it, and economic value from not being generally known. It does not turn on who or what created the information. So purely AI-generated code that cannot be copyrighted may still be protectable as a trade secret. That matters most for SaaS code that never leaves your servers and least for software you distribute. Two caveats, in our view. AI-accelerated development can erode the value of secrecy, because competitors can independently reach similar solutions faster than before. And pasting proprietary code into consumer AI tools may itself destroy trade secret status by disclosing it. One question is untested: the DTSA gives standing to an “owner” of the secret, and whether that maps cleanly onto AI-generated material has not been decided.

What’s Settled, What’s Not

The core rule is now settled: copyright protection requires meaningful human authorship, and with the Supreme Court declining to review Thaler, that requirement stands as U.S. law unless Congress changes it. What remains open is narrower, but still consequential. Fair use for AI training is unresolved; the Third Circuit heard argument in Thomson Reuters v. Ross in June 2026 and a ruling is pending, so how models may lawfully be built is not yet fixed. The other open question is where the line sits for AI-assisted works, meaning how much human creative input is enough to earn protection, which cases like Allen v. Perlmutter are now testing. For creators and companies, the safe posture is to document real human authorship and treat purely AI-generated portions as unprotected. Trade secret law can still protect such code where real secrecy is maintained. And the Ninth Circuit’s September 2026 ruling in Doe v. GitHub channels disputes over AI output into ordinary infringement and contract law rather than the DMCA.