The European Union will begin applying a new rule for companies that provide artificial intelligence systems within the bloc starting next Sunday, the 2nd. This requirement mandates that images, audio, and texts created by AI and designed to simulate authenticity must be properly marked with a digital label.
This imposition is part of the guidelines established in the European Union's Artificial Intelligence Act and takes immediate effect for any new systems launched on the European market. However, platforms and tools that already existed will have an additional four-month grace period to comply with the requirements.
The requirement is part of the implementation of the European Union's Artificial Intelligence Act, which stipulates that all AI-generated content intended to replicate real-life situations or materials with a realistic aspect must carry a digital mark. This identification can be done using the standard created by the European Union itself or equivalent systems developed by other organizations.
The new rules exclude content intended for private use, such as private communications between users. Furthermore, works that are unequivocally artistic, including satire and fiction clearly recognizable as such, are also exempt.
Companies introducing new AI systems into the European market must comply with these provisions starting August 2nd. For tools that were already available before this date, the legislation grants an additional four-month period for adaptation.
Member of the European Parliament Sergey Lagodinsky, who participated in the negotiations of the European Union's Artificial Intelligence Act, commented on the implementation of the measure, stating that its scope goes beyond mere consumer protection. He explained to The Guardian that the initiative aims to protect democracy by preserving the authenticity of facts on the internet.
Failure to comply with these requirements can result in fines of up to 3% of the total gross revenue of the companies responsible for the systems.
Boniface de Champris, head of artificial intelligence policy at the Computer and Communications Industry Association, observed that many consumers may be surprised to discover where these markings will start appearing. According to him, although social media already uses mechanisms to flag some AI-generated content, the trend is that these labels will also be visible in areas such as advertising, cinema, and the publishing market, where the technology is already widely applied.
The text also highlights that large digital platforms already have policies focused on identifying AI-generated content, although this practice is not always completely effective. Among the examples cited, TikTok requires creators to label AI-made material, but cases where this content goes unnoticed still occur, such as short videos featuring supposed doctors giving questionable health advice.
Additionally, corporations like Google and Meta already employ proprietary technologies to assist in identifying synthetic materials. Google announced that its SynthID tool has been used to tag over 100 billion images and an equivalent volume of audio amounting to approximately 60 thousand years of playback. Meta also maintains specific policies to identify photorealistic images created by artificial intelligence.
