The Communications Regulatory Authority (RRT) explains in which cases the obligation to label content applies to businesses, institutions, and content creators, what exceptions apply, and how the content should be labeled.
Read more Former Russian politician found dead in Warsaw
“The transparency requirements aim not to restrict the use of AI, but to ensure that a person can understand when they are interacting with AI, when certain AI systems are used in relation to them, and when the content they see or hear is created or modified using AI. This is important both for trust in AI technologies and for reducing the risks of misinformation, manipulation, or fraud,” says RRT Chair Jūratė Šovienė.
The requirements to label AI-created or modified content apply to companies, public sector institutions, media outlets, advertising creators, and other legal and natural persons using this content in professional or economic activities, e.g., opinion formers conducting commercial activities on social networks.
In cases where AI systems are used only for personal, non-professional purposes, the requirements to label content do not apply.
What content must be labeled
The AI Act provides for the obligation to label AI-generated or modified video, audio, or audiovisual content if it falls under the so-called synthetic media (deepfake) category. Such video or audio content very realistically imitates real people, objects, places, or events, which a person might mistakenly consider authentic or real.
It is also mandatory to label AI-created or modified text that has not been thoroughly reviewed by a person if it is published publicly and aims to inform the public on matters of public interest.
Public publication of text means that it can be accessible to an undefined, fairly large number of potential readers. Text published on social networks is public, except in cases where access is limited to a small closed group of people (e.g., small private chat groups) or it is published through an organization’s internal communication channels and similar.
Topics considered to be of public interest include politics, democratic processes, public administration and services, law enforcement, public safety and health, environmental protection, economy, finance, science, or culture – topics that may be important for public discussions.
When labeling is not required
The labeling requirement does not apply when AI-created text has been reviewed by a person who has verified the facts presented. Mere formal text checking, e.g., correcting spelling or punctuation errors, is not sufficient. Labeling also does not apply to AI-created or modified text that does not contain statements related to topics important to the public, such as health, consumer safety, sustainability, and similar (e.g., advertising or product descriptions).
In the case of deepfake, labeling is not required when AI is used only for auxiliary editing functions or making minor changes that do not alter a person’s perception of the authenticity or reality of such content. This may include, for example, editing background details, adjusting lighting, color correction, adapting audio parameters, noise reduction, or file compression.
Simplified labeling applies to deepfake content that is clearly part of artistic, creative, satirical, fictional, or similar works or programs. In such cases, labeling should not interfere with viewing or enjoying the work, but the obligation to label such content still remains.
How to label AI-created content
Labeling must be clear, easily noticeable, and immediately understandable. The icon must be clearly visible and recognizable from the first contact with the content, directly integrated into the deepfake or text, and remain visible when sharing or downloading the content.
RRT recommends using uniform EU icons in English, which can be downloaded for free.
Who will oversee compliance with the requirements
In Lithuania, the supervision of compliance with the transparency requirements of the AI Act is entrusted to the RRT.
RRT notes that initially it will focus mainly on prevention, informing and consulting representatives of the public and business to help target groups prepare for the application of the new requirements.
The AI Act provides for significant sanctions for failure to fulfill transparency duties – fines can reach up to 15 million euros or up to 3% of the company’s total global annual turnover. However, RRT experts emphasize that this is the maximum possible fine, applicable in exceptional circumstances and in cases of gross non-compliance with transparency obligations.
RRT experts also note that the AI Act provides special rules regarding the size of fines for small and medium-sized enterprises and startups, and the specific procedure for applying fines in Lithuania will be established in a law currently under consideration in the Seimas. It is planned to differentiate fine amounts based on the severity of the violation, the operator’s income, mitigating or aggravating circumstances. A warning is also provided as a possible sanction for minor violations.
More information about the requirements effective from August 2, 2026, can be found in the FAQ.
