In the European Union new rules ensuring the transparency of artificial intelligence (AI) usage have officially come into force . According to Euronews, this is one of the important parts of the phased implementation of the AI Act of the European Union.
From now on, companies developing and using AI systems are required to clearly label content created or significantly altered by artificial intelligence so that users can easily distinguish it from human-generated materials.
What do the new rules imply
The new regulations concern generative models, chatbots, as well as services that generate text, images, audio, and videos.
Immediate notification. Users must be informed if they interact with an AI system.
Mandatory labeling. AI-generated materials must have special markings (digital watermarks, metadata, and other technological means).
Combatting deepfakes. Special attention is paid to realistic fake images, videos, and audio recordings (deepfakes), the labeling of which becomes mandatory.
Monitoring public texts. Texts that represent public interest, if generated by AI and published without human/editorial intervention, are also subject to mandatory labeling.
Exceptions. The new requirements do not apply to personal use of AI. Exceptions are also provided for artistic, creative, satirical, and fictional works.
Fines and transition period
Companies face hefty financial penalties for non-compliance with the requirements.
However, existing systems have been granted a transition period to align their operations with the new norms.
Response from tech giants
Major tech companies are already taking steps to comply with the new requirements:
Meta has implemented the “AI Info” label on Facebook and Instagram.
TikTok has long required the indication of AI-generated media content.
Google, OpenAI, Nvidia, and Apple are working on developing cutting-edge technologies for digital content labeling.
In the European Union, it is emphasized that these steps are aimed at preventing the spread of misinformation and maintaining public trust in digital content.
