EU introduces tough regulations for AI-generated content
EU users are unlikely to notice AI labels right away (Photo: Unsplash)
New rules for AI systems will come into force across the European Union on August 2. From that date, chatbots will be required to inform users that they are interacting with AI, while deepfakes and other AI-generated content must carry clear labels identifying their artificial origin, according to a publication in the Official Journal of the European Union.
Key obligations
Under the new rules, AI providers and developers must comply with several transparency requirements:
System identification. Chatbots and voice assistants must be designed so users clearly understand they are communicating with an AI system.
Content labeling. Developers must embed technical markers, such as watermarks or metadata, into AI-generated images, audio, video, and text.
Deepfake disclosure. Any published AI-generated content depicting real people must clearly indicate that it was created artificially.
Public information content. AI-generated materials intended to inform the public about matters of public interest must be labeled if they have not undergone human review or if no individual has taken editorial responsibility for them.
Exceptions. Labeling requirements do not apply when AI is used solely for standard editing tasks, such as correcting spelling or grammar.
The rules also clarify that fictional content, such as images of people flying through the air, does not fall under the definition of deepfakes.
Why the labels may not be noticeable
Although the rules are now in force, most users are unlikely to immediately notice visible labels or major changes to AI-generated content.
First, Article 50(2) primarily requires machine-readable labeling, such as metadata or invisible watermarks that can be detected by specialized tools, rather than prominent visual icons.
The use of the European Commission's three official visual icons for fully AI-generated, partially AI-modified, or AI-assisted content remains voluntary under the Code of Practice.
Second, there is a transition period. Generative AI systems already on the market before August 2, 2026, have an additional four months to implement the required labeling measures, with a deadline of December 2, 2026.
This applies to most popular AI services currently available.
Fines and future restrictions
Violations of the transparency requirements under the AI Act can result in fines of up to €15 million or 3% of a company's global annual turnover.
The regulation also sets out a timetable for additional restrictions.
Starting December 2, 2026, it will become illegal in the EU to offer AI systems that generate non-consensual intimate images or child sexual abuse material.
Violations of these rules may result in the highest level of penalties, reaching €35 million or 7% of a company's annual global turnover.
Meanwhile, stricter rules governing high-risk AI systems have been postponed until December 2027 and August 2028.
Read also: EU introduces restrictions after declaring social media algorithms a threat to children
