*The Artificial Intelligence (“AI”) Act was approved by the members of the European Parliament on March 13, 2024. Although the AI Act does not include a direct regulation in regard to copyrights, it is seen that the obligations imposed on AI providers in Article 53 are intended to prevent copyright infringements.
Generative AI models such as ChatGPT, BERT, LaMDA, DALL-E2, which enable the creation of original content by processing data such as text, image and sound, are now becoming widely used.
With the increasingly intensive use of AI models, the need to fill the legal gap in this field has become ever more prominent. The necessity for an effective law to prevent risks arising particularly from AI models and to determine the obligations of AI providers has long been under discussion.1
Indeed, this matter holds significant importance on the agenda of the European Parliament in recent years. The spearhead AI Act, which was proposed by the European Commission in 2021, has begun to outline the obligations of AI providers, among other related issues.
Subsequently, the draft was submitted to the European Parliament (“EP”) by the relevant committee along with several proposed amendments and the Parliament approved its negotiating mandate on June 14, 2023. On December 09, 2023, Parliament and Council negotiators reached a provisional agreement on the final version of the AI Act.2 Eventually, the AI Act was approved by the members of the EP on March 13, 2024 with 523 votes in favor, 46 against and 49 abstentions.3 The AI Act will be implemented progressively and is expected to come into force fully in 2025-2026.4
In our opinion, the AI Act is of great importance as it is the first known legal regulation in this field worldwide. The main objectives of this Act are to improve market functioning within the European Union ("EU"), to promote the adoption of human-centered and reliable AI technology, to ensure a high degree of protection of fundamental rights, including health, security, democracy, the rule of law and environmental protection against the harmful effects of AI systems and to promote innovation.
So what does the AI Act entail in terms of copyrights?
It should be noted that the AI Act does not directly regulate copyright. The Act essentially establishes harmonized rules for the introduction, deployment and use of AI systems within the EU. It classifies AI technologies into various risk categories, prohibits certain types of AI applications and imposes special rules and obligations on operators of high-risk AI systems.
Article 53 of the AI Act which regulates the obligations of AI providers is of a nature that may be associated with copyright. In this context, providers of general-purpose AI models shall:
a) Prepare and keep up-to-date technical documentation of the AI model, including its training processes and provide this documentation to the AI Office6 and national competent authorities upon request;
b) Prepare, update and make available information and documentation for AI system providers who intend to integrate a general-purpose AI model into their own AI systems;
c) Provide right holders with the possibility to refuse to authorize use with reservations7, in order to ensure compliance with European Union copyright law and in accordance with Article 4(3) of Directive 2019/790 on Copyright and Related Rights in the Digital Single Market, which sets out exceptions and limitations to rights in relation to use for text and data mining purposes;
d) Prepare and make publicly available a sufficiently detailed summary about the content used for the training of the AI model, according to the template provided by the AI Office.8
It is seen that the obligations cited in subparagraphs (c) and (d), in particular, are in relation to copyrights. When these obligations are evaluated together, a framework is established in which AI providers will act in accordance with existing legal regulations by respecting and observing the copyrights of third parties, sharing with the public the data they use while training AI, and providing mechanisms9 whereby right holders can decide whether or not to allow the use.
Each of these issues aim to prevent possible infringements, to provide protection to right holders, and to provide AI providers with a more specific legal basis for their activities compared to before.
Even if they are not directly related to copyright, the Act also addresses transparency obligations. In this respect, Article 50 includes obligations such as clearly notifying the user of cases where AI interacts directly with individuals, marking that the content generated by AI, informing the individuals who are exposed to these processes in terms of systems that can perform operations (such as emotion recognition biometric categorization), and performing this in accordance with the procedures and laws, and explaining to users of an AI system that produces or manipulates image, audio or video content and creates deep fakes that the content is artificially generated or manipulated.
In particular, the last obligation regarding marking has an informative/notifying function, informing that the content was created with AI. We consider that this marking function may contribute to the decision-making of the relevant authorities in cases of potential disputes by considering this issue in cases of copyright ownership claims and/or infringement of rights regarding content.
In conclusion, it can be said that the Act in question is a regulation that provides specific, transparent mechanisms for the use of AI technologies, respecting the rights of third parties - including copyright rights- and deterring violations by imposing sanctions10. However, the adequacy of these regulations in meeting the needs will only become clear in praxis, and it is certain that legal regulations will continue to evolve within the scope of developments in the field.
Given that the relevant Act does not contain any regulations regarding the nature of the work generated by AI systems or potential copyright ownership and infringement cases related to them, it appears that, at least for now, clarification of these points remains with the courts.
1 https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
2 https://artificialintelligenceact.eu/developments/
3 https://www.europarl.europa.eu/news/en/press-room/20240308IPR19015/artificial-intelligence-act-meps-adopt-landmark-law
4 https://www.europarl.europa.eu/doceo/document/A-9-2023-0188-AM-808-808_EN.pdf
6 https://digital-strategy.ec.europa.eu/en/policies/ai-office The AI Office was established by the European Commission as a center of expertise on AI and to define a uniform management system within the Union.
7 https://enterprise.gov.ie/en/consultations/consultations-files/summary-articles-of-directive-eu-2019-790.pdf The relevant regulation establishes an obligation to allow the reproduction and extraction of legally accessible work for text and data mining purposes. However, if the right holders do not authorize the use, then this exception will not apply. Such reproduction may continue for as long as necessary for text and data mining purposes.
8 https://www.europarl.europa.eu/doceo/document/TA-9-2024-0138_EN.html
9 This issue is interpreted in doctrinal sources to correspond to "opt-out", i.e, mechanisms that express the right holder's determination that they do not consent and prohibit the use in a situation where the use is already considered permitted.
10 In Article 99 titled “Sanctions” of the relevant Act, among other provisions, in case of non-compliance with transparency obligations by AI providers, authorized representatives, importers, distributors, and other actors, administrative sanctions will be imposed in the amount equivalent to 15 million Euros. In the event of the violation being committed by a business entity, either the relevant amount or 3% of its total annual worldwide turnover for the preceding financial year of the violating entity, whichever is higher, will be applied. In case of provision of incorrect, incomplete, or misleading information, administrative sanctions will be imposed in the amount equivalent to 7.5 million Euros. And if the violation is committed by a business entity, either the relevant amount or 1% of the total annual worldwide turnover of the preceding financial year of the violating entity, whichever is higher, will be applied.