Copyright and Ownership Rights in the Era of Generative AI

When is an AI output a copyrighted work

Under the provisions of Section 2(1) of Act No. 121/2000 Coll., the Copyright Act (hereinafter referred to as the "Copyright Act"), under Czech legislation, a work is considered to be a literary work, other artistic work, and scientific work which is the unique result of the creative activity of the author and is expressed in any objectively perceivable form, including electronic form (hereinafter referred to as the "work"). Furthermore, pursuant to Section 5(1) of the Copyright Act, the author is the natural person who created the work.

Therefore, a copyrighted work must meet certain conditions set by the Copyright Act in order to be considered as such. Based on this definition, we can conclude that any output created by AI cannot be considered a copyrighted work, particularly because the work will not be created by a natural person, but by machine artificial intelligence, and also because it will not be a unique result of creative activity.

User liability for AI inputs and outputs

Currently, two AI softwares are most frequently discussed, namely ChatGPT and Midjourney.

The ChatGPT software was created by OpenAI, LLC (hereinafter "OpenAI"), which is a software generating text content, while the Midjourney software was created by a group of developers and deals with generating images based on text.

So how do the individual softwares work? ChatGPT generates output based on input from the user, and together they form the generated content. OpenAI states in its Terms of Use (hereinafter "Terms of Use") that users themselves bear responsibility both for their inputs, i.e., the prompts they enter into ChatGPT, and for the use of the generated output. For inputs uploaded to ChatGPT, they must verify that they do not violate any law, including the Copyright Act. This applies particularly to the Midjourney software, in cases where users upload an image to the software that may be subject to copyright. Users also bear responsibility for the overall content, including the output from ChatGPT and Midjourney, where they must ensure that by publishing such generated content, they do not infringe the copyright of another person whose work the AI software used for its output.

Given that AI software does not cite the sources it drew from for its output, it is necessary to watch out for obvious plagiarism. This is especially true considering the volume of sources AI works with, and in the case of visual content generated by Midjourney, such plagiarism can be easily detectable.

When copyright can arise for modified output

In the next section, we will focus on more practical aspects related to outputs from AI software, particularly the ownership rights to outputs from ChatGPT. As mentioned above, copyright under the Copyright Act generally cannot be held for generative AI outputs, although there is a certain exception worth mentioning. This is represented by a situation where the user modifies the AI output, enriches the output with their own unique creative elements, and only then presents such output as their copyrighted work. In this case, through the creative processing of the AI output, the user may acquire copyright to a so-called derivative work, which is created by the creative processing of another work under Section 2(4) of the Copyright Act. However, it must be kept in mind that in this case, it is necessary to process the AI output and add one's own thoughts and ideas to it.

Ownership rights to AI outputs and provider terms

However, an important question remains as to who can use this content to their advantage and how. In the Terms of Use, OpenAI states that the user retains ownership rights to their input and also acquires ownership rights to the output from ChatGPT. The user can then use this output for commercial purposes as well. The developers of Midjourney have set their rules differently, where ownership rights apply only to images generated in the paid version of the software, while free accounts can only generate images for non-commercial use.

Therefore, the user of the AI output does not acquire copyright to the output, but does acquire ownership rights to it. In practice, this means that the user can dispose of the output in accordance with the Terms of Use and the law. Although they become the owner of the generated content, there is no guarantee that another user will not use a similar or even the same output, perhaps even for the same purpose, given that the output cannot be protected as a copyrighted work. In the Terms of Use, OpenAI also points this out, stating that the output from ChatGPT may not be unique and other users may receive a similar or the same output. In this, we can see another aspect of why AI output cannot be considered a copyrighted work, as the fact that there can be multiple identical outputs goes against the characteristic features of a copyrighted work, where its uniqueness is expected and considered one of the fundamental conditions for the creation of a copyrighted work. If a user publicly uses content purely generated by AI without any of their own creative additions and finds that someone else has used the same content, they cannot seek any protection, as this content will not be subject to copyright protection.

Furthermore, attention should also be drawn to OpenAI's Sharing and Publication Policy, which states that the user should publish the work under their own name and at the same time inform that the content was created by AI. Although OpenAI is supposed to act in a way that does not infringe anyone's copyright, in the event of an infringement of copyright or other intellectual property rights of third parties, OpenAI disclaims any liability. Currently, a number of copyright lawsuits have been filed against OpenAI, so it can be expected that courts will soon begin to address and shape this issue.

Risk matrix for using generative AI

Final summary

In conclusion, it can be summarized that the AI user will acquire ownership rights to the output, but this will be independent of copyright, which they will not acquire as it will not be a copyrighted work. Using AI-generated outputs can make work easier for users in many respects, but when using content generated by AI, care must always be taken to ensure that this content is used in accordance with the terms of use of the respective developers and applicable laws, and overall, one must be cautious when using AI outputs.