
Law as of 4 November 2026. Author: Marek Stolorz, attorney-at-law (radca prawny).
Content generated entirely by artificial intelligence is not a work under Polish law, and nobody holds copyright in it. It is different when a human makes their own creative contribution: selecting, arranging, reworking and deciding on the final shape. That person is then the author of their contribution. At the same time, AI output may infringe other people’s copyright, and since August 2026 the EU AI Act requires certain content to be labelled.
Is AI output a copyright work?
The Polish Copyright Act protects any “manifestation of creative activity of an individual nature” (Article 1(1)), and copyright belongs to the creator (Article 8(1)). Both concepts refer to a human being: creative activity means choices that someone makes consciously. Likewise, in Infopaq (C-5/08) the Court of Justice of the EU held that copyright protects only original subject matter, meaning the author’s own intellectual creation (paragraph 37). In a text, that creativity lies in the choice, sequence and combination of words (paragraph 45), that is, in decisions made by a human.
An AI model is not an author, and typing a prompt is usually not enough to make a person the author of the output. An image generated from a single prompt is therefore most often in the public domain: anyone may use it, including your competitors.
U.S. courts have reached the same conclusion. In Thaler v. Perlmutter, the court of appeals held on 18 March 2025 that copyright requires a human author, and on 2 March 2026 the U.S. Supreme Court declined to hear the case. The court stressed that this does not rule out protection for works made with the help of AI where a human is the author.
When is the person using AI the author?
When AI is a tool and the human makes the creative decisions. Examples: a designer who generates elements and then composes and reworks the illustration himself; a writer who uses AI for a draft but writes and edits the text herself. What is protected is the human contribution, not what the machine added.
In practice, it is worth documenting the process: keep successive versions, prompts and edits. In a dispute, this is evidence that the work came from a human’s creative choices.
Can AI infringe someone else’s copyright?
Yes. AI output may reproduce protected elements of existing works: passages of text, a distinctive composition, a character. The fact that the output is not itself a work does not mean you are free to publish it. If it contains someone else’s work, distributing it may infringe the author’s rights, and liability lies with whoever uses it. So before using an AI-generated image or text in advertising or on a website, check that it does not resemble a known work.
Training models is a separate issue. Since 20 September 2024, Polish law allows published works to be reproduced for text and data mining, unless the rightholder has reserved this (Article 26³ of the Copyright Act). For works available online, the reservation must be machine-readable. The AI Act requires providers of general-purpose AI models to respect such reservations and to publish a summary of the content used for training (Article 53(1)(c) and (d)).
What the AI Act changes: labelling AI content
Article 50 of the AI Act applies from 2 August 2026. Regulation (EU) 2026/1744, which postponed other AI Act deadlines, kept that date, but gave providers of systems placed on the market before 2 August 2026 until 2 December 2026 to mark their output. The key obligations:
| Who | Obligation | From |
|---|---|---|
| Provider of an AI system generating images, audio, video or text | mark the output in a machine-readable format as AI-generated (Article 50(2)) | 2.08.2026; for systems placed on the market earlier, 2.12.2026 |
| Provider of a chatbot | inform people that they are interacting with AI, unless this is obvious (Article 50(1)) | 2.08.2026 |
| Business publishing a deepfake | disclose that the content was generated or manipulated (Article 50(4)) | 2.08.2026 |
| Publisher of AI-generated text on matters of public interest | disclose that the text was AI-generated, unless it has undergone human editorial review and someone holds editorial responsibility (Article 50(4)) | 2.08.2026 |
For most businesses this means: if you publish AI-generated images or recordings showing real people or events, you must disclose it. Texts that have been edited and for which someone takes responsibility are not covered by the disclosure obligation.
Tool terms and contracts
Rights in the output are also governed by the tool’s terms of use. Providers take different approaches to commercial use and to whether they assign their rights to the user, “if any”. Check the current terms before using AI content in your business.
If you commission images, text or code from an agency or freelancer, agree in the contract whether they may use AI and whether they warrant that the result does not infringe third-party rights. Without such a clause, you may pay for something in which nobody holds copyright, so nothing can be transferred to you.
Frequently asked questions
Is an image from Midjourney protected by copyright?
If it was created from a prompt alone, usually not. Protection covers only the human creative contribution, such as composition and edits.
Can I use AI content commercially?
Copyright law does not prohibit it, but check the tool’s terms and make sure the output does not reproduce someone else’s work.
Do I have to label AI-generated content?
Deepfakes must be labelled from 2 August 2026. Texts edited by a person who takes responsibility for them do not need labelling.
Can I prohibit AI training on my texts?
Yes. You must reserve this expressly, and for online content in a machine-readable format (Article 26³ of the Polish Copyright Act).
We advise on IT contracts, licences and copyright transfers: see New technology law and Contracts.



