EU publishes icons for labelling AI-generated content**

The European Commission has published a set of EU icons for labelling certain AI-generated or manipulated content.

The icons are intended for creators, publishers and other deployers of generative AI systems. They may be used to help people recognise when content has been artificially generated or altered.

The icons are linked to Article 50(4) of the EU AI Act. This provision requires disclosure for certain AI-generated or manipulated content, including deepfakes and AI-generated text, published to inform the public on matters of public interest.

A deepfake is an AI-generated or manipulated image, audio or video content that resembles real persons, objects, places, entities or events and could falsely appear authentic or truthful.

The disclosure rule also covers AI-generated or manipulated text on matters of public interest when the text has not gone through human review or editorial control and no person or organisation has taken editorial responsibility for it.

The Commission says not all AI-generated content must be labelled. There are limitations and exceptions. For artistic, creative, satirical, fictional or similar works, disclosure obligations are limited so that they do not interfere with the display or enjoyment of the work.

The obligation also does not apply where the use of deepfakes or relevant AI-generated text is authorised by law for detecting, preventing, investigating or prosecuting criminal offences.

For text, the disclosure obligation does not apply where the content has undergone human review or editorial control and where a natural or legal person holds editorial responsibility for publication.

The EU icon set includes several options. A basic icon can be used when AI was involved in creating deepfake content or relevant published text, or where a custom text label or interactive second layer is used. Other icons distinguish between fully AI-generated content and content that has been partly modified using AI.

The Commission says the icons were user-tested. It found that recognition improved when the basic icon was accompanied by a text label, such as ‘modified’.

Use of the EU icons is optional. However, the labelling duties under the AI Act are not optional where they apply. The Commission also states that using the icons does not by itself prove legal compliance. Deployers remain responsible for ensuring that disclosures comply with the AI Act’s requirements.

The icons should be clearly visible when a person first sees the content. They should be placed where they are not hidden by overlays and should generally be embedded into the content or shown through an equivalent user interface method.

The Commission also encourages accessibility measures. Labels should use plain language, avoid unnecessary jargon, and be readable by assistive technologies where possible.

The icons are freely available for public use without attribution to the Commission or the AI Office. Signatories of the Code of Practice on marking and labelling AI-generated content are expected to follow the relevant placement specifications.