Türkiye follows the European Union in the field of artificial intelligence regulations, as in KVKK and E-commerce laws. It entered into force in the EU on 1 August 2024. EU AI Act’i model alarak hazırlanmış gibi görünen Yapay Zeka Yasa TeklifiIt was presented to the parliament on June 1, 2024. In this article, we will summarize the differences we find important between the proposal in Turkey and the EU AI Act, and what the gray zones and opportunities that may arise with this law may be.

Let's Compare the EU AI Act and the Artificial Intelligence Bill in Turkey

First of all, we should mention that the EU AI Act has a long history. This law, the draft of which started in 2021, took its final form in the last year with articles aimed at ensuring the security of these models and ensuring that they do not go beyond people's control and expectations, as general-purpose artificial intelligence technologies such as advanced language models (LLMs: Large Language Models) have become widespread. Let us underline that the bill in Turkey is a proposal, not a draft, so it is reasonable to expect that it has not yet been developed as long and detailed as a law. Its length is 10% of the EU AI Act, and its content and details are just as simple and less comprehensive. Still, we can say that this proposal is a good start in terms of determining certain principles.

Prohibited Artificial Intelligence Application Areas

The EU AI Act strictly prohibits the use of artificial intelligence in the following areas:

1. Deceiving people, trying to mislead their behavior, exploiting people's weaknesses
2. Collecting people's sensitive data without their consent and causing prejudice and discrimination with this data
3. Developing social scoring systems that will impact people's access to services and job opportunities
4. Real-time identification and surveillance and visual collection of people in public spaces without their knowledge or consent
5. Detecting people's emotions in education and business

Our bill states that artificial intelligence must comply with the following principles, but does not set certain limits on its application areas:

• Security: Safe and risk-free operation of artificial intelligence systems
• Transparency: Artificial intelligence systems and processes should be clear and understandable
• fairness: Artificial intelligence systems should not discriminate and be fair
• Accountability: Identifying and auditing individuals and institutions responsible for artificial intelligence results
• Security: Artificial intelligence systems comply with the principles of protecting personal data and keeping it confidential.

Although the proposal states that some application areas, such as autonomous vehicles and medical diagnostic systems, will be considered high-risk systems and will be subject to stricter rules, it does not contain application bans that offer as sharp lines as the EU AI Act and center on citizen rights. The aspect of this difference that raises the most questions is that it is not clear how much flexibility some practices that are clearly banned in the EU will be welcomed in Turkey, as we have an example like China, which keeps its citizens under strict surveillance with artificial intelligence technology. Even though it is still at the proposal stage, the fact that such a basic sanction is not included in the proposal raises curiosity as to whether it will be evaluated at the draft stage. On the other hand, we can of course predict that these five principles listed will play an important role in preventing risks arising from artificial intelligence. These principles are the principles announced for the year 2021-2025. In the National Artificial Intelligence Strategy yerli yapay zeka ekosistemimizin takip etmesi hedeflenen prensiplerle de neredeyse tamamen örtüşüyor. Bu açıdan da bu yeni yasa teklifinin mevcut milli yapay zeka stratejisini güçlendiren ve destekleyen nitelikte bir teklif olduğunu söyleyebiliriz.

Definitions of Actors in the Artificial Intelligence Ecosystem

A second difference is that artificial intelligence actors are defined more clearly in the EU AI Act than in the bill in Turkey. Since criminal sanctions in the law are made directly based on these definitions, clarity in these definitions is also very important.

EU AI Act’in A 68-item definitions section there is. There is a wide list of definitions, from basic concepts such as artificial intelligence and risk to niche concepts such as artificial intelligence model distributor and artificial intelligence model importer. The bill in Turkey has made the definitions of artificial intelligence, provider, distributor / user, importer, distributor and artificial intelligence operator, which we do not use much in daily life but are directly related to the law, in general parallel to the definitions in the EU AI Act.

These definitions appear perhaps in one of the most striking sections of the proposal, where it is explained what responsibilities and penalties are deemed appropriate for these defined actors. The EU AI Act opens a separate title for distributors, model developers, model distributors and many other actors in the section where responsibilities and penalties are listed. In the bill proposal in Turkey, the penalties are specified to cover all actors in the ecosystem, called artificial intelligence operators, whose definition we have shared verbatim below.

Artificial Intelligence Operators: This definition will cover all stakeholders in the artificial intelligence ecosystem

has been kept broadly. Providers, distributors, users, importers and distributors

This definition, which includes all of it, describes the responsibilities and responsibilities of all parties within the scope of the law.

plays a central role in determining its obligations.

For example, according to the proposal, it is not clear yet whether the fine, which is defined as a violation of obligations and may reach up to 15 million TL or 3% of the organization's annual turnover, will be applied to a company that develops a model that does not comply with the obligations, to a user who has put the model into use without checking whether this model complies with the obligations, or perhaps to a distributor in Turkey who offers this model developed abroad. There is no doubt that these penalties will be a deterrent factor, but as they are in the current text, these penalties may cause concern due to the actors in the artificial intelligence ecosystem not clearly understanding their responsibilities.

Advanced General Artificial Intelligence

The third and last point we will touch upon is that general artificial intelligence systems are included in the EU AI Act. an important place olmasına rağmen bizdeki teklifte bu modellere değinilmemiş olması. Bunlar, ChatGPT’nin arka planındaki gibi çok gelişmiş ve çok amaçlı, sadece teknik geliştirme basamakları dahi milyonlarca dolara mal olan, üretilmeleri için kullanılan bilgisayar gücünün de belli bir miktardan yüksek olduğu modeller. Mevcut modeller henüz insanlığı tehlikeye sokabilecek potansiyelde görülmese de bu modellerin ne kadar hızlı bir şekilde yeni kapabiliteler kazanbildiğini gören yapay zeka uzmanları, çok daha gelişmiş ve riskli modellerin yakında üretileceğini öngörüyor ve bu modellerin güvenilir, transparan,  insan değerleriyle ve niyetleriyle uyumlu olmasının çok büyük önem arz ettiğinin altını çiziyor. Bu alanda uluslararası işbirliği sağlamak amacıyla başlatılan AI Safety Summit serisine Türkiye de katılmıştı, Paris’te düzenlenecek bir sonraki zirveye de katılacağımızı öngörebiliriz. EU AI Act’te tanımlanan gelişmişlik düzeyinde modelleri üretebilen OpenAI, DeepMind gibi şirketler şimdilik sayılı. Türkiye’de bu seviyede modeller üreten özel veya milli bir girişim olup olmadığına dair de net bir bilgimiz yok. Dolayısıyla yasanın bu bölümünün, Türkiye ile çok bir ilgisi olduğunu söyleyemeyiz. Fakat yapay zekanın artık milli güvenlik ve ekonomik kalkınma planlarına girdiği günümüzde, Türkiye’nin gelişmiş yapay zeka modellerine karşı ne gibi bir politika izleyeceğini, Türkiye’deki şirketlerin bu alanda ne gibi adımlar atacağını da bilmiyoruz. Bu gelişmiş modellerin denetlenmesi, test edilmesi, bu modellerin kullanıldığı uygulamaların kontrol altında tutulması oldukça büyük bir denetleme ekosistemini beraberinde getirecek. Bu alanlarla ilgiliyseniz daha çok bilgi almak için bizimle iletişime geçebilirsiniz.

What Awaits Us?

If the artificial intelligence bill in Turkey is accepted, most of the details at the application level will probably be defined later. This also applies to the EU AI Act. In light of this bill, the EU established two organizations called European AI Board and EU AI Office. Among these two, operational tasks such as implementation of laws, definition of relevant tests, supervision of compliance with laws and coordination of countries seem to be more under the responsibility of the EU AI Office.  Undoubtedly, all of these processes will lead to the emergence of many new actors in the private sector in the EU and to change the business processes of existing actors. Our predictions for actors in Turkey are as follows:

• If you want to establish an auditing company within the framework of this law in Turkey, provide consultancy to the regulation of this law, or follow the legislation closely as a lawyer, the EU AI Office will probably be one of the centers kept under the radar while defining the detailed practices in Turkey.
• If you are an organization that produces or uses artificial intelligence, you will be interested in the application of the law to you rather than the making process. In this case, three different actors come to the fore:
  • Organizations that produce or use artificial intelligence
  • Entrepreneurs or existing actors who want to establish an inspection business in this field
  • Actors that may be indirectly concerned with practices in this field, such as insurance and occupational safety

Most companies that will undergo a significant audit under this law will need some form of consultancy or product to prepare before the audit. While they may establish this internally, they may also want to receive external consultancy or a solution. In addition, it is possible to foresee that the institutions that will officially audit this legal regulation will be private auditing organizations that have undergone some kind of accreditation. Therefore, this bill could lead to many new initiatives in the field of supervision. Likewise, we cannot say that it is too early for fields such as insurance and occupational safety, which indirectly interact with these fields, to follow this bill closely and prepare what products they will offer in response to the demands from companies to protect themselves against this law. 

What about those to be audited? Probably the greatest uncertainty awaits them. Both the risk of facing high penalties and the current uncertainty in the draft law may cause uneasiness in companies, and may even be interpreted from an investment perspective as a move that will inhibit innovation and R&D investments of companies in the field of artificial intelligence. On the other hand, for the sustainable growth of the technology ecosystem, innovations must be capable of protecting, improving and acting in harmony with people and societies. From this perspective, we can expect that such practices, which may create uncertainty and turbulence in the short term, will benefit every actor in an order where they are defined in a clear, informed manner and implemented fairly.

Of course, when examining this bill, it is necessary to keep in mind that this text is only a proposal, and that the very specific language used in the law sometimes has a narrowing effect on the law. In this journey from proposal to draft and acceptance, many details such as detailed definitions, which institution will maintain and supervise which legislation will be determined. There is no doubt that this proposal, which we consider to be a good start, will require more titles than the current ones, especially over time. For example, if the harmonization process with the EU comes to the agenda again, it will be important to clearly state high-risk and prohibited practices in our law, as in the EU AI Act. Making the definitions of artificial intelligence actors and the responsibilities of each of them much more detailed than the current bill will support the reliability of the legal processes that will operate within the framework of this law. We will continue to follow the process closely and share it with you, hoping that such positive changes will be included in the later stages of the bill.

Resources

We will include important developments regarding the EU AI Act. our monthly newsletter and edited by Risto Uuk, focusing solely on the EU AI Act this newsletter We recommend you follow.