The European Union will begin enforcing new transparency rules on Sunday requiring companies to clearly label artificial intelligence-generated content, marking another phase in the bloc’s wide-ranging AI legislation aimed at tackling misinformation and improving public trust online.
Under the new rules, businesses must ensure that AI systems, including chatbots, clearly identify themselves as artificial intelligence tools. Images, text, audio and other content created using AI must also carry labels indicating that they were generated or altered by the technology. Companies can meet these requirements by using digital watermarks and similar identification methods that allow AI-generated material to be easily detected.
Businesses that fail to comply with the regulations could face substantial financial penalties.
European officials say the measures are designed to help people distinguish between authentic and AI-generated material as advances in the technology make it increasingly difficult to tell the difference.
“Generative AI enables disinformation to be created at unprecedented scale, tailored to specific audiences, and disseminated with remarkable speed,” an EU official said.
The official added that the new rules are intended to protect public confidence in the information people see, hear and read online, particularly as synthetic content becomes more convincing.
A major focus of the regulations is deepfakes, including AI-generated or manipulated videos, images, audio recordings and text that closely resemble real people or events. The EU considers such content a growing risk because it can be used to spread false information and mislead the public.
The requirements apply to content produced for professional purposes. Individuals using AI solely for personal activities are exempt from the rules. Text created by AI to inform the public on matters of general interest must also be labelled if it is published without meaningful human editorial oversight.
Companies operating existing AI systems have until December 2 to bring their services into compliance. The legislation also provides exemptions for artistic, creative, satirical and fictional works.
The rules have drawn criticism from some businesses, which argue that implementing the requirements across rapidly expanding AI services will be difficult. Ashley Casovan of the International Association of Privacy Professionals acknowledged the challenges but said compliance has often appeared difficult before becoming standard practice.
Several major technology companies have already introduced AI labels ahead of the regulations. TikTok has required creators to identify AI-generated images, videos and audio for years and says more than three billion pieces of content have already been labelled through creator disclosures and automated detection tools.
Meta has added its “AI Info” label to qualifying posts on Facebook and Instagram, while Google has signed the EU’s AI transparency code of conduct and is working with Nvidia, OpenAI and Apple on digital content tagging technology.
Google’s Karen Massin, however, cautioned that increasingly complex regulations could create confusion if users are presented with too many overlapping labels and legal notices instead of clear, easy-to-understand information.
