The rise of artificial intelligence tools like Yiaho, ChatGPT, and DALL-E raises an increasingly debated question: are AI-generated creations plagiarism?
The answer, in short, is that it depends on the circumstances:
- If an AI uses existing works without permission to generate content that too closely resembles copyrighted works, there could be a rights violation.
- However, if the AI creates sufficiently original content without direct copying, it cannot always be considered plagiarism.
In the absence of specific regulations in many countries, including France, current copyright laws are sometimes vague regarding the use of AI. These creations are therefore in a legal “gray area.”
In some cases, works created by AI cannot even benefit from copyright protection because they are not considered “human.”
Let’s explore in more detail what legal texts say in France and internationally:
How AI works: between inspiration and reproduction
Generative AI systems, such as Yiaho, ChatGPT, or DALL-E, are trained on massive databases containing human works (texts, images, videos, etc.). These tools do not make direct “copies,” but analyze and reproduce patterns or structures to generate new content.
If this generation is too close to a pre-existing work, it can raise plagiarism concerns.
The essential question is therefore whether the AI reproduces an existing work in a recognizable way (which would be plagiarism), or if it produces something sufficiently different and original to be considered an autonomous creation.
Test our free online AI detector
The legal framework: copyright and AI in France and around the world
In France: the Intellectual Property Code
In France, the Intellectual Property Code (CPI) governs copyright. It protects any “work of the mind” from its creation, as stated in Article L.111-1: “The author of a work of the mind enjoys an exclusive incorporeal property right in this work, enforceable against all.” This includes texts, images, music, and other artistic creations.
The problem with AI is that its operation relies on the massive use of data and pre-existing works, often without the creators’ permission. Here are some key points:
- Use of works to train AI: French law does not yet provide a specific framework for the use of protected works by AI for training purposes. In principle, a protected work cannot be used without explicit permission, unless it benefits from an exception (such as the short quotation exception, Article L.122-5). However, AI models are not limited to partial quotes, which makes their use contentious.
- AI-generated works: According to Article L.112-1 of the CPI, only works showing human originality are protected. A creation generated by an AI, being produced by a machine without direct human intervention, could not, in theory, be protected by copyright.
See also: Here are the legal professions that can be replaced by AI
In Europe: the Copyright Directive and the text and data mining exception
The Directive (EU) 2019/790 on copyright in the Digital Single Market, adopted by the European Union, offers a framework for the use of data for research and development purposes, notably with the text and data mining exception. This allows the extraction of large amounts of information (texts, images, etc.) to train AI, provided these uses are intended for scientific research and not for commercial purposes, unless the author of the work explicitly objects.
- Protection of AI-generated works: As in France, European legislation does not recognize works created exclusively by machines as original. They cannot therefore be protected by copyright, unless significant human intervention is demonstrated.
In the United States: the Copyright Act and AI
In the United States, the Copyright Act (federal copyright law) operates on the same principle as in France and Europe: only works created by humans can be protected. The United States Copyright Office confirmed in 2022 and 2023 that works generated by AI, such as those produced by MidJourney or ChatGPT, cannot benefit from copyright protection unless there is substantial human intervention.
This raises questions regarding the rights of users or companies using AI tools to generate content: if an AI-generated work is not protected, what prevents someone else from reproducing it?
Read also: An AI to learn English: Here is the trick
International legislation: WIPO and the Berne Convention
On an international level, the Berne Convention (1971), overseen by the World Intellectual Property Organization (WIPO), remains the reference for copyright. This convention only recognizes human creations as being eligible for copyright protection.
However, WIPO has recently launched debates on the impact of AI on artistic and intellectual creation, but no rules have yet been adopted to regulate AI-generated works.
Plagiarism risks for creators
One of the main concerns for artists, writers, and creators is the use of their protected works to train AI without their consent. This could be considered a form of plagiarism or copyright infringement, especially if the generated work strongly resembles the original work.
Discover more on this: How to create drawings and images with artificial intelligence
Lawsuits have already been launched by creators of images and visual works, such as photographers, against AI companies, alleging the illegal use of their works to train algorithms. They are seeking either compensation for the use of their works or restrictions on the use of their data.
Is ChatGPT detectable?
One concern for teachers, publishers, and professionals is the ability to detect if a text was generated by ChatGPT or another AI. Currently, detection tools exist, such as our Yiaho AI detector tool or Copyleaks, which have developed algorithms capable of identifying AI-generated texts. These tools look for patterns specific to machine-produced texts, such as overly generic phrases or syntactic structures that lack natural variation.
However, these tools are not infallible, and the detection of AI-generated text remains a developing field. The more ChatGPT is able to produce texts that mimic the human way of writing, the harder it becomes to differentiate them from those written by humans.
In France, no specific law yet regulates the detection of AI-generated texts, but the use of these technologies is starting to spread, particularly in the educational sector, to prevent cheating.
Is ChatGPT considered plagiarism?
ChatGPT, as a generative artificial intelligence tool, is not inherently a plagiarism tool. It works by analyzing huge sets of textual data (such as books, articles, and websites) to create new and coherent responses.
However, this does not mean it is immune to generating content that may be similar, or even too close, to existing texts. If ChatGPT faithfully reproduces a copyrighted work without citing the source or without consent, it can be considered a form of plagiarism.
However, OpenAI, the company behind ChatGPT, has put measures in place to avoid this by limiting its ability to generate texts that are direct copies. ChatGPT doesn’t just memorize and reproduce; it combines information to produce new text.
Despite this, in certain situations, it can generate phrases or expressions close to pre-existing works, which can pose a legal problem if the work is protected by copyright.
Read also: Tip: Here is the best AI for students
Do AIs commit plagiarism?
AIs, like ChatGPT or DALL-E, do not directly commit acts of plagiarism in the sense that they do not consciously “know” what they are doing. However, they use existing data, sometimes protected, for training, and can generate content that closely resembles human works.
In France, for example, such use could be considered a violation of the Intellectual Property Code if the generated work uses substantial elements of an original work without permission.
In the case of DALL-E, for example, questions have been raised about the fact that some generated images looked too much like works by photographers or illustrators, even if these resemblances were not exact copies. Thus, although AI does not necessarily commit plagiarism per se, risks of copyright infringement exist.
Is ChatGPT cheating?
The use of ChatGPT can be perceived as cheating in certain contexts, particularly in school or university settings, where the production of original work is required. If a student uses ChatGPT to write an essay or assignment without mentioning it, it could be considered an act of academic fraud. Educational institutions, in France and elsewhere, are implementing strict policies to regulate the use of AI in assignments.
In the professional world, it depends on the context. Using ChatGPT to generate ideas, write drafts, or automate tasks is not necessarily cheating, provided the use of the tool is transparent and accepted by the company or clients.
However, if a person uses ChatGPT to generate work that is supposed to be personal and original without informing their employer or clients, it could be considered dishonest.
The key here lies in transparency. The use of tools like ChatGPT must be clearly defined and accepted within the rules established by the institution or company.
See on this topic: 5 reasons why AI cannot replace human intelligence
AI and plagiarism: Possible solutions and adjustments
To address these concerns, several legal solutions are being studied:
- Specific licenses for AI training: Licenses could allow companies to pay to legally use protected works for training AI models, thus ensuring compensation for creators.
- Remuneration for artists and creators: A mechanism similar to neighboring rights for performers could be considered, offering remuneration to creators whose works are used to train AI.
- New rules for AI-generated works: Legislators could consider laws that would define how AI-generated works can be protected or not, and how to avoid potential cases of plagiarism.
Conclusion
The question of whether AI constitutes plagiarism is still widely debated, with legal gray areas. In the absence of clear texts adapted to new technologies, AI-generated works fall into a copyright gray area.
However, with the rapid evolution of these technologies, it is likely that laws will soon clarify these situations. In the meantime, creators must remain vigilant about the use of their works by AI systems and ensure that their rights are respected.
