Introduction of New AI Transparency Rules in the EU

Starting August 2, new requirements for the transparency of artificial intelligence usage will come into effect in the European Union, as outlined in Article 50 of the AI Act. These rules directly impact users and mandate the clear identification of chatbots as machines, as well as the disclosure of deepfakes. Regulation (EU) 2026/1744 was published on July 24 and will take effect on July 27. The European Commission released clarifications on the application of Article 50 less than two weeks before the deadline.

Main Requirements of Article 50

Providers are required to design chatbots and voice assistants in a way that makes it clear to users they are interacting with a machine. Outputs generated by AI must be labeled in a machine-readable format, meaning technical tags should be added to images, audio, video, and text. Anyone publishing a deepfake must disclose this if the content realistically depicts real people. Similar requirements apply to AI-generated text published to inform the public on matters of societal interest, unless the text has been reviewed by a human editor responsible for its content.

Exceptions to the Rules

If AI is used solely for routine editing, such as correcting spelling or grammar, labeling is not required. Content that is clearly fictional also does not fall under the definition of a deepfake. For example, an image of people flying through the air would be exempt.

Labeling Features

A machine-readable tag differs from a visible one: Article 50(2) refers to watermarks and metadata that can be detected by specialized tools, rather than visual icons. For generative systems already on the market before August 2, there is a four-month transition period to implement labeling, until December 2, 2026. This applies to nearly all modern services.

Visible icons also exist: three options are proposed—for content created with AI involvement, fully generated, or partially modified. Their use is voluntary and part of the AI Content Transparency Code of Practice. However, having an icon does not exempt providers from fulfilling the requirements of Article 50.

Oversight and Penalties

In Germany, the Federal Network Agency has been appointed as the central market oversight authority, and the implementation law has already passed the Bundesrat. The agency is establishing a contact point for reporting potential violations. Non-compliance with transparency requirements can result in fines of up to 15 million euros or 3% of global annual turnover. The strictness of these rules for major providers can be seen in ongoing disputes over AI assistants on smartphones.

Additional Bans and Deadlines

From December 2, 2026, it will be prohibited in the EU to offer AI systems that create intimate images without consent or materials depicting sexual violence against children. The maximum fines for such violations are up to 35 million euros or 7% of annual turnover. Stricter rules for high-risk AI will come into force later—in December 2027 and August 2028.