EU AI Act: Regulating the Unseen
The last two years have been very exciting; this period has been marked by discussions not only on the possibilities of artificial intelligence (AI) but also on addressing its real dangers and identifying ways to manage them. While two years may seem like a short time, the response to these challenges has been surprisingly swift. The European Union (EU) emerged as the first in the world to develop a risk-based regulatory law on AI, with the Artificial Intelligence Act (AI Act) coming into effect on August 1, 2024.
As part of this initiative, a critical regulation prohibiting the use of systems posing unacceptable risks took effect on February 2, 2025. Previous attempts to control AI, including the EU’s ethical guidelines established in 2019, aimed to ensure the safe, ethical, and reliable development and application of AI technologies. However, these guidelines often attempted to confine technology within ethical and moral boundaries, which, in essence, was an impossible mission.
Rethinking Ethics
It is not the ethics of AI that needs rethinking, but rather our own human ethics and morality, particularly in light of AI’s existence. The law’s mission is to raise public awareness about artificial intelligence and foster a kind of “digital consciousness.” Understanding the risks associated with AI services or products enables individuals to better grasp the opportunities and threats linked to AI. The goal is not merely to shape technology to our needs but to establish rules that encourage conscious and risk-based usage, ultimately protecting society from potential dangers.
AI should not be feared, as it is not an apocalyptic entity. Instead, concern should be directed towards individuals who intend to misuse AI technologies.
Future-Proof Regulatory Framework
The EU AI Act not only addresses current AI challenges but also provides a future-proof regulatory framework capable of adapting to technological advancements. Its primary objective is to build and maintain trust in AI, safeguarding transparency and accountability.
In the future, AI-based solutions will be classified into four categories based on their risk levels: unacceptable, high, limited, and minimal risk. Systems posing unacceptable risks will be banned outright. For the other categories, proportional measures must be applied, such as ensuring transparency through operational reports, maintaining human oversight, and implementing stricter testing for higher-risk applications. For minimal-risk classifications, informing users that they are interacting with AI will suffice, and any AI-generated content must be clearly labeled.
Examples of AI Risk Levels
To illustrate these categories, consider a Call Centre Customer Service System:
- Minimal-risk: An AI assistant entertaining callers while they wait.
- Limited-risk: An AI that assists in optimizing customer service processes, requiring clear disclosure that the customer is communicating with AI.
- High-risk: AI systems utilizing automated decision-making or emotion detection, necessitating stricter protocols.
- Unacceptable risk: AI used for covert surveillance or manipulation will be banned.
Companies implementing such systems must prioritize transparency, inform users about lawful usage, and select solutions that comply with EU law.
Global Considerations
While regulating AI usage within the EU is commendable, questions arise regarding the global landscape. In countries like China, AI systems deemed unacceptable by the EU are utilized at the state level, such as the social credit system, which monitors citizens’ actions through surveillance technologies and financial transactions. This system allocates points based on behavior, impacting creditworthiness and access to social benefits.
The EU AI Act prohibits practices such as social scoring and real-time biometric identification in public spaces. For instance, a shopping mall should not be equipped with systems capable of identifying individuals. While general data collection, such as age or gender, is permissible for service improvement, the storage and analysis of personal data are strictly regulated.
Manipulative Practices and Emotional Detection
Another banned practice is the use of AI systems that manipulate cognitive behavior. For example, tactics like “Last item available!” or “Three people are currently viewing this product!” create a false sense of urgency, pressuring consumers into purchasing decisions. These emotionally targeted ads exploit psychological vulnerabilities to influence choices.
One concerning practice deemed unacceptable by the law is the use of hidden emotion detection in workplaces and educational institutions. Employees do not want to be monitored by software analyzing their emotions during video calls, assessing their stress or satisfaction levels. Such practices infringe upon fundamental rights and perpetuate pressure to exhibit ‘positive’ emotions.
The Future of AI and Human Interaction
In the future, individuals may migrate not for better weather or living standards but to evade invasive state monitoring of their private lives. In this regard, the EU may represent a safer place to live. Companies violating the AI Act could face fines of up to 35 million euros or 7% of their previous year’s global turnover, whichever is higher. Similar to the GDPR, this regulatory framework carries a significant deterrent effect.
Ultimately, the greatest benefit of the AI Act is empowering individuals to choose how much access and influence AI should have in their lives and the extent to which they share their thoughts with AI technologies.
Can We Trust AI?
The real question is not whether we can trust AI but whether we can afford not to trust it. As society navigates the complexities of AI, it becomes essential to learn how to coexist with this technology. AI awareness will evolve into a crucial skill for both companies and individuals.
AI sensitivity encompasses ethical, environmental, and social dimensions. It involves ensuring responsible AI use aligned with human ethical principles, avoiding unnecessary digital waste, and understanding AI’s societal impact on the labor market and various social groups. As we prepare for the societal consequences of AI, including workplace transformations, it is clear that we are entering an exciting new era.
Over the past two years, major AI solutions have emerged rapidly, and increasingly sophisticated multimodal AI systems that combine text, image, voice, and video are now on the rise. The combined effect of these innovations will soon sweep across the world, and society must be prepared.
