Kto własnością jest obraz wygenerowany przez sztuczną inteligencję
Stephen Thaler spędził lata, arguing that his “Creativity Machine” deserved authorship credit for the artwork it generated. On March 2, 2026, the US Supreme Court ended that argument by declining to hear it, leaving intact a ruling with direct consequences for anyone building products on generative AI: copyright belongs to humans, and the line is now set for the foreseeable future.
The US Copyright Office has held a consistent position: works created solely by AI, without human creative input, are not eligible for copyright protection. That position traces back to a long-standing legal tradition. The Copyright Clause of the US Constitution, read alongside cases such as Burrow-Giles Lithographic Co. v. Sarony (1884), defines authorship as the “original intellectual conception of the author.” The D.C. Circuit Court of Appeals, affirming Judge Beryl Howell’s ruling in Thaler’s 2025 appeal, described human authorship as the “backbone of copyright law.”
AI tools, on that reading, occupy the same legal position as a camera or a word processor: useful instruments, not creators. Any copyright in an AI-assisted work attaches only to what the human contributed.
The Zarya of the Dawn registration decision from 2023 is the clearest illustration of how the Office applies this in practice. Author Kris Kashtanova received copyright for the text and her human selection and arrangement of the graphic novel’s elements, but the Office denied protection for the individual AI-generated images. The deciding factor was “genuine creative control”: selecting, arranging, editing or meaningfully modifying AI output in a way that reflects original human expression. Rewriting AI-generated lyrics, re-recording a melody, or composing a new arrangement from AI-generated fragments can all clear that threshold.
The EU AI Act, in force since August 1, 2024, approaches the problem from a different angle. It does not define authorship or copyrightability for AI-generated content, but it places compliance obligations on the companies building the models. Providers of general-purpose AI (GPAI) models must demonstrate compliance with EU copyright law and publish detailed summaries of the data used to train their systems. Fines for non-compliance can reach up to 3% of global annual turnover or 15 million euros, and these provisions become enforceable on August 2, 2026.
The legal landscape is further complicated by a ruling from Munich. On July 31, 2026, the Landgericht München I decided the first European case on copyright liability for a generative AI music tool, GEMA v. Suno Inc. The court prohibited four acts with respect to six GEMA-administered compositions: reproduction for training purposes in the United States, retention of those works within the AI model in Germany, communication to the public by making the trained model available online, and reproduction and communication to the public through AI-generated output tracks. The court applied US copyright law to the US training acts and rejected Suno’s fair use defense, finding the training use commercial and substitutive.
The ruling creates liability exposure for AI music generation companies that train models on recordings administered by territorial collecting societies in EU Member States without securing prior licenses. The court’s reasoning was significant: it found that a work memorized in an AI model constitutes a reproduction under copyright law, comparing the model to a progressively stored JPEG file in which information is scattered.
Different jurisdictions are taking different paths. In China, the Beijing Internet Court ruled in September 2025 that AI-generated images are eligible for copyright protection, but the creator must document their creative thinking, input commands, and the selection/modification process with supporting evidence. Japan broadly permits AI training under Article 30-4 of its copyright law, unless it unreasonably prejudices rights holders. The United Kingdom, after its March 18, 2026 government report, dropped its proposed text-and-data-mining exception with opt-out and reverted to relying on existing copyright law plus industry-led licensing.
The practical takeaway for creators, developers, and policymakers is clear: the law has not caught up to the technology, and it may not for years. In the meantime, the safest strategy is to assume that anything created purely by AI has no copyright protection, and that the human contribution must be substantial, documented, and original to claim ownership. The question is no longer whether AI can create, but whether the law recognizes who did.
Oznaczenie treści AI
Ten powstał z udziałem sztucznej inteligencji: research oraz tekst przygotował i zredagował agent AI. Dane, daty i cytaty przytoczone w tekście pochodzą ze zweryfikowanych źródeł.
