AI-Generated and AI-Assisted Code Under EU Law: Intellectual Property and Compliance Challenges
Part 2: Terms of Service of AI platforms & EU main regulations for computer programs
Av. Mirela Takacs · April 20, 2026 · 11 min read
The governing terms of AI platforms
When a generative artificial intelligence (GenAI) system is used in the development of a software program, it is also necessary to examine the terms of use of such a tool. For example, for the most commonly used AI systems, whether it is the Terms of Service (consumer[1] or commercial[2]) of Claude or those of Open AI for ChatGPT[3], as applicable in EU, they all state that the providers assign to their customer all their right, title, and interest (if any) in Outputs. Google Services on the other hand says that Google won’t claim ownership over that content[4].
Copyright under Directive 2009/24/EC
As regards the legislation, at the European Union level, computer programs are protected by copyright as literary works. According to Article 1 and Article 2 of Directive 2009/24/EC[5]:
- protection shall apply to the expression in any form of a computer program (it is protected only their materialization through source code, object code, preparatory design material provided that the nature of the preparatory work is such that a computer program can result from it at a later stage[6]);
- ideas and principles which underlie any element of a computer program, including those which underlie its interfaces, are not protected by copyright;
- a computer program shall be protected if it is original, in the sense that it represents an intellectual creation of the author. No other criteria shall be applied to determine its eligibility for protection;
- the author of a computer program shall be the natural person or group of natural persons who created the program or, if the legislation of the Member State permits, the legal person designated as the rightholder by that legislation; in respect of a computer program created by a group of natural persons jointly, the exclusive rights shall be owned jointly; where a computer program is created by an employee in the execution of his duties or following the instructions given by his employer, the employer exclusively shall be entitled to exercise all economic rights in the program so created, unless otherwise provided by contract.
In contrast, Directive 2001/29/EC[7] (the ordinary law of copyright) does not define the notion of author nor does it provide for the possibility of designating a legal person as the rightholder, as in the case of Directive 2009/24/EC.
Patents under European Patent Convention
Another method of protection is through patents, but in Europe the protection of computer programs through patents is a complex subject, with distinctions between what is and is not patentable.
According to art. 52(2)(c) and (3) of the European Patent Convention, computer programs are excluded from patentability if claimed as such. However, following the generally applicable criteria for Art. 52(2) and (3) (G‑II, 2), the exclusion does not apply to computer programs with technical character. Technical character in a computer program depends on the production of a “further technical effect” when run on a computer[8]. Once technical character is established, novelty and inventive step are then evaluated.
In other words, while copyright protects the expression of a computer program, patent law protects the technical solution, namely the program’s functionality, methods or processes if they produce a technical effect when implemented and executed on a computer program.
Legal protection of the AI output
All these generative AI systems define output as the content generated and returned by the service based on user input. However, in order to address the issue of assignment or ownership of rights, we should first analyze whether the output – the result generated by the platform – benefits from legal protection.
Thus, considering the legal conditions regarding the protection of computer programs listed above, in order for the code generated by a Generative AI system to be protected by copyright, it must meet the condition of originality, that is, it must be the expression of an intellectual creation of the author.
In relation to computer programs, Directive 2009/24/EC lays down the general principle that the author must be a natural person or group of natural persons, therefore human, except where Member States allow by their legislation that a legal person may be designated as the rightholder of a computer program. Distinct from this aspect, the Directive also regulates the ownership of economic rights in the context of an employment relationship.
Therefore, since the protection of computer programs under copyright is reserved for human intellectual creation, the notion of authorship being grounded on the idea of human creativity, the possibilty of a non-human entity, such as an AI tool, being recognized as author of an output would be excluded because it cannot produce originality in the legal sense (being irrelevant whether a machine is capable of autonomous creative behavior[9]) and the AI system is not a human author and does not have legal personality.
Consequently, it is also considered that the output generated by AI with insignificant human involvement at any stage of the development (conception, execution or redaction phases), such as through simple instructions, prompts or without making any significant changes to the result that reflect free creative choices, would not benefit from copyright protection, the legal requirements not being met. Human creativity is relevant for analyzing copyright protection[10].
In such cases, the AI output would, in principle, fall outside the scope of copyright protection and there are no related rights available specifically for software code as such. Still, as seen above, another form of protection would be through patents. But, with regard to inventorship, the European Patent Office, in its decisions, rejected patent applications naming an AI system as the inventor on the grounds that an inventor designated in the application has to be a human being, and not a machine[11].
Another possible legal regime from the intellectual property framework would be trade secret law considering that it does not provide for the requirement of human creation, but it is necessary that the information is not known among or readily accessible to persons within the circles that normally deal with the kind of information in question[12]. And in this context, the AI system’s terms of service regarding the user input become particularly relevant.
One might think that the output could be used freely without asking anyone’s permission, but, paradoxically, its use may still infringe third-party rights (for details, see Part 3).
On the other hand, if the output is modified or adjusted, whether directly by the user or by an AI agent acting on user’s instructions and within the limits of the permissions granted to it, at any stage of the development, it is considered that the AI system serves as an execution assistant that facilitates or accelerates tasks that would otherwise require much longer work.
For example, AI agents for coding and DevOps can perform tasks such as interpreting a task description, generating code, running tests, debugging, commiting to repository, creating pull requests, deploying to staging, and monitoring CI/CD pipelines. In these scenarios, they can have access to external systems such as Git repositories, CI/CD pipelines, cloud infrastructure, IDEs, terminals, or package managers[13].
In such cases, for the final result to benefit from copyright protection under Directive 2009/24/EC it is considered necessary to demonstrate a creative human process throughout development that can be verified, a creative control, and that the author’s creative choices are identifiable. However, this raised the question if there is a minimum level of creative human involvement sufficient to meet the legal requirements and whether such a threshold should be regulated[14].
It should be noted that, at this moment, there are no official rules at EU level regarding copyright over AI-generated or AI-assisted outputs. The above conclusions are based on current scholarship and CJEU case law on the concept of originality[15], also applicable to computer programs. However, as clarified in Painer (Case C‑145/10, p. 88-92), originality results from the author making various free and creative choices in several ways and at various points in its production, thereby expressing his personality and stamping the work with his “personal touch” by making those various choices.
In this sense, some studies note that even mere human intervention at the conception and redaction stages could suffice for copyright protection, even if the execution belongs entirely to the AI system, by analogy with the situation in which an author conceives (“masterminds”) and supervises a work executed by others without contributing materially to the execution[16].
Until further developments, at present the evaluation for copyright protection of code produced with/by AI has to be made on a case-by-case basis, taking into account the specific circumstances of each situation, and resolved through the application of existing legal frameworks and principles.
As for the developers of the AI model, they would not, in principle, be considered entitled to rights in the output, unless they played a creative role in the process of producing a specific result[17]. As long as the developers of the AI system did not make a specific creative effort for that specific output, and their work was limited to building the system that generally generates outputs, they would not be entitled to authorship for a specific work. Even if they were entitled to invoke such rights, usually the company operating the AI system grants them to the user through the applicable Terms of Service, as seen above.
Returning to Chardet, if a similar case were to be resolved under EU law, the maintainer could hold copyright over the new version of the program to the extent that he demonstrates his creative choices throughout development.
For example, what we know from Blanchard’s explanations for the rewriting of Chardet 7.0.0 is that he wrote the requirements list, configured permissions, gave instructions, authorized exceptions, verified the output and published everything, all within 5 days, compared to what would have otherwise taken months. So clearly some work was done, his involvement being far from insignificant in terms of time and effort, but what is relevant is to what extent his contributions reflect creative choices.
Therefore, it is possible that the new version of the library benefits from protection under the law. But another problem now arises, as the original author also mentioned in his request to revert Chardet to its original license – the fact that it was not a clean room implementation. What relevance would this have? Let’s continue in Part 3.
Part 1: A real-world example: the relicensing of Chardet
Part 4: CJEU case law on computer programs
Part 5: Literal Elements vs. Non-Literal Elements in Computer Programs
[1] https://www.anthropic.com/legal/consumer-terms
[2] https://www.anthropic.com/legal/commercial-terms
[3] https://openai.com/policies/eu-terms-of-use/
[4] https://policies.google.com/terms?hl=en-US
[5] DIRECTIVE 2009/24/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 23 April 2009 on the legal protection of computer programs, OJ L 111, 5 May 2009
[6] Recital 7 of DIRECTIVE 2009/24/EC
[7] Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the harmonisation of certain aspects of copyright and related rights in the information society, OJ L 167
[8] https://www.epo.org/en/legal/guidelines-epc/2025/g_ii_3_6.html
[9] C. HARTMANN and others, Trends and Developments in Artificial Intelligence, Challenges to the Intellectual Property Rights Framework, p. 70, 76, 80, 84
[10] Trends and Developments in Artificial Intelligence, Challenges to the Intellectual Property Rights Framework, p. 84-85
[11] https://www.epo.org/en/news-events/news/epo-publishes-grounds-its-decision-refuse-two-patent-applications-naming-machine
[12] The Development of Generative Artificial Intelligence from a Copyright Perspective (EUIPO, 2025), p. 33
[13] L. Nannini and others. AI AGENTS UNDER EU LAW, https://arxiv.org/abs/2604.04604
[14] N. Lucchi, Serra Hunter, Generative AI and Copyright – Training, Creation, Regulation, Section 4.4
[15] Infopaq International, Case C-5/08, p. 35 and 45; Painer, Case C-145/10, p. 87-92; Cofemel, Case C‑683/17, p. 30 and 3; Konektra, Case C‑795/23, p. 50;
[16] Trends and Developments in Artificial Intelligence, Challenges to the Intellectual Property Rights Framework, p. 82
[17] Trends and Developments in Artificial Intelligence, Challenges to the Intellectual Property Rights Framework, p. 85
The content of this website is provided for informational purposes only and does not constitute legal advice. The law in this area is subject to ongoing developments. Please consult a qualified lawyer for advice specific to your situation.