Abstract

The transparency of communication with a machine is one of the foundations of the AI Act. This duty applies to all AI systems that interact directly with humans, unless communication with the AI system should be obvious to the user. A part of the broader transparency requirement is machine-readable labelling of content generated by artificial intelligence. The fulfilment of this obligation is of particular importance primarily for the creative industry or in the area of litigation, which is an important element of identifying the origin of the content in dispute. The act also imposes an obligation to identify deepfake understood as artificially generated content resembling or even mimicking the elements that make up our reality. The establishment of this obligation is intended to contribute to a more effective fight against the spread of disinformation.
The wording of individual obligations and their scope is far from ideal, and therefore it will be necessary to develop appropriate standards so that the goal of introducing the principle of transparency is actually achieved.