AI—but how?

Primarily to protect against the risks associated with the use of AI, “Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonized rules on artificial intelligence,” commonly referred to as the “AI Act,” was adopted on June 13, 2024. It is an extremely comprehensive piece of legislation comprising 180 recitals, 113 articles, and 13 annexes. The regulation is set to take effect on August 2, 2026; however, certain chapters and articles are to be applied earlier, in several phases, starting on February 2, 2025. Among other things, the regulation addresses the risk of dependence on and the spread of misinformation, as well as the risk of artificial intelligence being used for purposes that threaten human existence, such as the development of chemical weapons.

According to the regulations, an AI “operator”— that is, the person who uses an AI system under their own responsibility (Art. 3 No. 4 AI-Act), must indicate when publishing a text that the text was generated by AI (Art. 50(4) AI-Act). An exception applies if the text has been reviewed by a human and if a natural or legal person bears editorial responsibility for the publication. For image and audio recordings, the regulations are stricter in that it does not matter whether they are to be published. Employers who use an AI system are required under Art. 4 of the AI-Act—regardless of company size—to ensure that their employees possess a sufficient level of AI literacy and are encouraged to offer training for this purpose as needed. The use of AI is completely prohibited, for example, under Article 5(1)(c) of the AI Act, for so-called “social scoring,” in which individuals are evaluated based on their behavior and personality for the purpose of differential treatment.

Compliance with the regulations is to be monitored by national authorities (Art. 70 of the AI Act); in Germany, discussions are currently underway between the state data protection authorities and the Federal Network Agency regarding this matter. At the EU level, new bodies are being established (such as the so-called AI Board) to support national authorities and contribute their expertise to the further development of the rules. It remains unclear at this time what legal consequences may result from non-compliance with the regulations. Pursuant to Article 99(1) of the AI Act, member states are required to establish provisions for sanctions by the time the regulation enters into force. These may include fines of up to 35 million euros or seven percent of the previous year’s revenue.

While the U.S. appears to be moving away from regulations of all kinds, the EU will have to address the question of to what extent a balance can be struck between protection against abuse and freedom for innovation. Critics of the European approach see this as a massive competitive disadvantage that will further exacerbate the EU’s lag in the technology market. It remains to be seen whether the EU will yield to competitive pressure from the U.S. and China or stick to its strict regulations in favor of fundamental rights and ethical standards.