The European Union's Role as a Super Regulator: The Implications of the AI Act for Companies in Turkey
1. Introduction
Artificial intelligence ("AI") systems have rapidly spread across all areas of life in recent years, from business to public services, security to entertainment. While this rapid development presents significant opportunities, it also raises concerns such as lack of oversight, lack of transparency, and social security. The European Union ("EU") has taken an ambitious step in technology law history with the European Union Artificial Intelligence Act ("AI Act/Law"), the world's first comprehensive artificial intelligence law, to manage these risks.[1]
The main goal of the AI Act is to classify AI applications based on a risk-based approach and impose stricter obligations as the potential level of harm increases. This framework aims to both open the door to technological innovation and protect the fundamental rights of individuals. In addition, the economic, social, and technological impacts that the law will have in the short and medium term are also being intensely debated.
The effects of the law extend beyond the EU's borders. The global regulatory influence known in academia as the "Brussels Effect" shows that regulations adopted in the EU often become international standards.[2] The EU's strong consumer market and regulatory reputation compel other countries to adopt EU standards and encourage global companies to apply the same standards worldwide. Therefore, the AI Act will have direct consequences for countries trading with the EU, including Turkey.
2. Legal Framework
The AI Act is considered the first comprehensive legal regulation on artificial intelligence. The law addresses the development and use of artificial intelligence with a risk-based approach, imposing different obligations according to the potential harm it creates. In this context, while some artificial intelligence applications are completely banned, others are subject to strict requirements such as transparency, risk management, and human oversight. Thus, the AI Act establishes different levels of control and accountability mechanisms according to different risk levels.
These regulations will enter into force according to a phased schedule[3] :
- July 12, 2024 – Publication of the law in the Official Journal of the European Union,
- August 1, 2024 – Entry into force,
- February 2, 2025 – Prohibited applications come into effect,
- August 2, 2025 – Entry into force of rules for general-purpose artificial intelligence systems,
- August 2, 2026 – Implementation of most obligations for high-risk systems.
Despite compliance issues that arose during the law's adoption process, the European Commission clearly stated that it would not back down from its timeline, sending a message of "no delay, no exemption"[4] . This stance demonstrates that the EU positions the Artificial Intelligence Act not merely as a technical regulation, but as an indispensable framework for social security and the protection of fundamental rights. The law is not limited to developers but covers different actors in the artificial intelligence value chain. Providers, implementers, importers, product manufacturers, and authorized representatives are regulated within this scope.[5] Two key actors in this value chain are artificial intelligence providers and implementers. Determining whether an organization is a provider or an implementer is critical, as these roles carry different obligations under the Act.[6]
- Providers are individuals or organizations that develop an artificial intelligence system or a general-purpose artificial intelligence ("GPAI") model, or facilitate its development on their behalf, and market it under their own name or trademark. For example, while a foundational model is considered GPAI, a chatbot or generative AI tool built on top of this model is classified as an "artificial intelligence system."
- Deployers are individuals or organizations that provide services to third parties using artificial intelligence systems. For example, companies that use a third-party chatbot for customer service are considered deployers.
- Importers are persons or organizations established in the EU that bring AI systems developed outside the EU into the EU market.
One of the key features of the law is its cross-border impact. If an artificial intelligence system or its output is used within the EU, providers and deployers outside the EU also fall within the scope of this regulation. For example, if a company in the EU sends data to an AI provider outside the EU for processing and this output is used in the EU, the relevant provider will be subject to the provisions of the AI Act. Therefore, companies operating outside the EU and providing services in the EU are required to appoint authorized representatives in the EU to carry out compliance processes.
Another prominent and noteworthy feature of the AI Act is its classification of artificial intelligence systems according to risk levels. This classification allows for stricter obligations to be imposed on systems with high potential harm, while lighter rules apply to low-risk applications.
- Minimal Risk: Applications that pose low risk in terms of rights or security. For example, a spam filter falls under the minimal risk category. There are no restrictions or mandatory obligations for artificial intelligence systems in this category. However, it is still recommended to follow general principles such as human oversight and preventing discrimination.
- Limited Risk: Covers AI systems that interact directly with users and can influence their choices. Chatbots and virtual assistants are examples. Transparency is required for these systems; users must know they are interacting with AI.
- High Risk: Covers all artificial intelligence applications that have the potential to harm people's health, safety, fundamental rights, or the environment. These systems must meet certain requirements to be placed on the EU market. Applications falling into the high-risk category are grouped into two main groups:
- Safety components of products subject to existing regulations: Artificial intelligence systems integrated into products such as medical devices, elevators, vehicles, or machines.
- [7] e independent artificial intelligence systems listed in Annex III:
- Biometric systems: Especially remote biometric identification and emotion recognition (highly controversial due to their direct impact on privacy and fundamental rights).
- Employment and workforce management: Artificial intelligence used in recruitment, performance evaluation, and task allocation (critical due to risks of inequality and discrimination).
- Law enforcement applications: Evidence analysis, crime prediction, or security risk assessment (a sensitive issue in terms of justice and human rights).
- Justice and democratic processes: Systems that support legal decisions or could influence elections (there are concerns that these could pose a direct threat to democracy).
- Unacceptable Risk: Artificial intelligence applications that pose a direct threat to fundamental rights and democracy fall into this category and are completely prohibited.
- Subliminal manipulation: Changing a person's behavior without their awareness. Example: Systems that encourage voting for a political party.
- Exploitation of vulnerabilities: Harmful applications based on age, health, or socio-economic status, such as toys that encourage dangerous behavior in children.
- General purpose social scoring: Rating individuals based on their social behavior, shopping, or social media interactions. This can lead to unfair job or credit rejections.
- Real-time remote biometric identification: Facial recognition systems in public spaces. Only limited exceptions are granted to law enforcement with judicial approval and for the purpose of preventing specific crimes.
This structure embodies the fundamental purpose of the AI Act: obligations increase as the level of risk increases. The aim is to protect individuals' fundamental rights while supporting innovation within safe boundaries.
Another comprehensive obligation under the AI Act concerns GPAI models. The initial draft of the law did not mention GPAI systems such as OpenAI or Aleph Alpha. However, during negotiations, the EU updated its proposal on this matter. The rules introduced for these models, including foundational models and generative AI tools, are primarily based on transparency. The EU AI Act distinguishes between two risk classes based on the processing power required to train the model: non-systemic risk and systemic risk. While all foundational models must meet transparency requirements, additional obligations have been introduced for GPAIs posing systemic risk. Systemic risk is defined as a model's potential to cause serious or foreseeable negative impacts on public safety, fundamental rights, security, or society at large. For example, models trained on very large datasets and requiring high processing power are considered systemic risk; models such as ChatGPT or Gemini, trained on billions of pieces of data and requiring massive supercomputers to run, are considered systemic risk. This is because such models can be used not only in one area but in many different tasks and, if misused, can have widespread negative effects on society. Publicly available open-source models may be exempt from stricter requirements if their licenses grant access, use, modification, and distribution rights to the model and its parameters. However, this exemption only applies if the model is not associated with high-risk or prohibited applications and does not carry a risk of manipulation.
The regulation's risk-based approach aims to protect fundamental rights and public safety while ensuring that innovative activities continue within predictable and controllable limits by imposing differentiated obligations on the development and use of artificial intelligence technologies.
3. AI Act Compliance Process for Companies and Its Implications for Turkey
As stated, the Act is a regulation that extends beyond EU borders and has a global impact. Non-EU providers offering AI systems or GPAI models to the EU market or putting them into service within the EU also fall within the scope of this regulation. Turkish companies will be directly affected by the geographical scope defined in Article 2 of the AI Act. Therefore, if an artificial intelligence system developed in Turkey is accessible to users in the EU or its outputs are used in the EU, it is directly subject to the provisions of the AI Act. Although there is no artificial intelligence law in force in Turkey yet, due to close commercial and technological relations with the EU, compliance with this regulation is not only a requirement for Turkish companies but also critical in terms of market sustainability and competitive advantage.
The first step for companies that need to comply with the AI Act requirements is to document their AI systems according to their characteristics and risk classes and classify these systems into "minimal, limited, high" risk categories. This classification allows companies to see in advance which systems will be subject to heavy obligations. Next, a comprehensive compliance plan must be prepared for each high-risk system: the plan should include technical documentation of the model, human oversight mechanisms, and transparency and accountability processes. Another important step is to determine the company's role; in this context, it is necessary to determine whether the company is a 'provider' or an 'implementer'. In the final stage, it becomes possible for companies to comply with the regulations by determining which obligations must be complied with.
- Conclusion
In terms of both its scope and impact, the law is not just a regulation, but a turning point that reshapes technology policies on a global scale. Thanks to its risk-based approach, risky applications are separated, aiming to protect individuals' fundamental rights and ensure innovation continues within a secure framework. The extension of this framework beyond EU borders through the "Brussels Effect" makes the compliance process inevitable for countries like Turkey, which have intensive commercial and technological ties with the EU. From a corporate perspective, the registration of AI systems, their classification according to risk categories, and the clarification of roles and responsibilities have now become a necessity. Although all these elements may bring additional costs and operational burdens in the short term, in the long term, they will provide reliability, transparency, and competitive advantage, forming the basis of a sustainable artificial intelligence ecosystem both within the EU and on a global scale.
[2] https://www.economist.com/europe/2021/04/24/the-eu-wants-to-become-the-worlds-super-regulator-in-ai
[4] https://www.reuters.com/world/europe/artificial-intelligence-rules-go-ahead-no-pause-eu-commission-says-2025-07-04/