On August 2, new requirements came into force in the European Union, requiring AI content to be labeled and users to be informed about interactions with AI systems.
This is part of the so-called EU AI Act, the first comprehensive law in the EU to regulate artificial intelligence.
From now on, providers and developers of artificial intelligence systems must clearly and visibly label content created with the help of AI. This applies, for example, to images, audio and video content that resembles real people, objects, places, entities or events, as well as texts that are “published for the purpose of informing the public on matters of public interest, without prior verification or editorial control”.
To this end, the European Union has developed special labels that can be used to identify AI content.
Exceptions are provided for AI systems that perform auxiliary functions, such as spell-checkers or photo-correction tools. An exception is also made for content that is “clearly artistic, creative, satirical or fictional in nature.” In such cases, the law allows other methods of information that “do not interfere with the viewing or perception of the work.”
Users must be warned about interaction with AI
Separate requirements apply to systems with which people directly interact. Users must clearly understand when they are communicating with an artificial intelligence rather than a human being — for example, with a chatbot, AI agent or digital avatar.
The European Commission has also published recommendations for developers and users of AI systems on how to comply with the new requirements.
What are the penalties for violations
The labeling obligations apply to both AI providers and users who use AI-generated content for professional or commercial purposes. These include companies, media outlets, professional influencers, authorities and civil society organizations.
European authorities will have the right to impose fines for violations of the requirements: up to 15 million euros, or 3% of annual global turnover, for companies, up to 750 thousand euros for EU institutions, bodies and agencies, and fines “taking into account the principle of proportionality for small and medium-sized enterprises”.
As noted, the law creates uniform rules for the use of artificial intelligence in the EU, aimed at developing safe and reliable AI technologies. At the same time, the document should reduce risks to human health, security, fundamental rights, democracy and the rule of law.