New transparency obligations regarding artificial intelligence (AI) systems in the European Union (EU) will begin to be implemented starting August 2.
elchi reports that the EU’s “Artificial Intelligence Act,” which entered into force on August 1, 2024, establishes different rules for AI systems based on their risk level. According to the law, certain AI applications that pose an unacceptable risk are prohibited, while strict obligations are imposed on high-risk systems.
In this Law, where rules regarding artificial intelligence come into force in stages, August 2 of this year is considered one of the most important implementation dates. From that date, the application of transparency obligations will begin.
The EU Commission and market surveillance authorities of member states will be able to effectively exercise their powers to inspect and impose sanctions.
The purpose of the new rules is to ensure that citizens can understand when they are interacting with artificial intelligence and whether the content they encounter has been created by artificial intelligence. This aims to prevent the deceptive use of AI, increase transparency in the information ecosystem, and strengthen trust in the digital environment.
The EU Commission has also published guidelines on the implementation of transparency obligations under the “Artificial Intelligence Act.”
The guidelines explain in which cases developers of AI systems and the organizations using them must inform users within the framework of the obligations that will come into force on August 2 of this year.
Accordingly, chatbots, AI agents, and similar interactive AI systems must clearly inform users that they are communicating directly with artificial intelligence.
In addition, photo, video, audio, and certain text content created or altered by artificial intelligence will be tagged with special machine-readable markers that allow them to be identified as AI-generated.
“Deepfake” content created through artificial intelligence must be clearly labeled as artificially generated.
It will also be required to notify users about AI-originated content that is of public interest and has not been verified by humans or undergone editorial control.
Organizations using AI systems that perform emotion recognition or biometric classification will also have obligations to clearly inform people that they are under the influence of such systems.
The guidelines emphasize that standard editing operations, such as spelling and grammar corrections, are not covered by these obligations.
On the other hand, most of the rules under the “Artificial Intelligence Act” will come into force on August 2, and the EU Commission and market surveillance authorities of member states will be able to exercise their inspection and sanction powers.
For AI systems placed on the market before August 2, a transition period until December 2 of this year will be granted to fulfill the technical obligations regarding the labeling and detectability of AI-generated content.