Starting August 2, the European Union will enforce new provisions of the Artificial Intelligence Act requiring companies to label deepfakes and other content created or altered by AI, infohub.kz reports.
The requirements apply both to AI system developers and organizations that professionally use these technologies. Providers of generative models must implement machine-readable labels indicating the artificial origin of images, audio recordings, videos, or text. Users must also be informed if they are interacting with a chatbot or other AI-based service.
Stricter rules apply to deepfakes—realistic images, audio, or video that mimic the appearance, voice, or actions of real people. Their creators and distributors must clearly indicate that the content is AI-generated or altered. Similar rules apply to AI-generated texts on socially significant topics published without full editorial oversight. Lighter requirements may apply to artistic, satirical, humorous, or fictional works, but the artificial origin must still be disclosed appropriately.
The EU notes that generative technology advances are making fake images, videos, and audio increasingly realistic. AI content can be used for creative purposes but also for spreading disinformation, political manipulation, and creating non-consensual explicit material featuring real individuals. An example is a 2022 fake video of Ukrainian President Volodymyr Zelenskyy allegedly calling on soldiers to lay down arms. In 2024, AI-generated explicit photos of singer Taylor Swift appeared online without her consent.
Experts doubt the new rules' practical effectiveness. One issue is the lack of a unified technical standard: companies use different labeling methods such as watermarks, metadata, and provenance tracking systems that are not always compatible. Digital labels can also be lost during editing, compression, or transmission across platforms. Another challenge is that internet content knows no borders: a deepfake created outside the EU can quickly spread to European users. The European Commission argues that technology alone is insufficient and recommends companies use multiple labeling and provenance-checking methods simultaneously for more reliable AI content detection.
Major AI developers, including OpenAI and Google, have supported the European transparency requirements and joined a voluntary code of practice. However, industry representatives warn that content generation and detection technologies evolve rapidly, so regulations will need frequent updates. The code itself is voluntary, but compliance with Article 50 of the EU AI Act is mandatory for all companies operating in the European market.