Image AI labels

Reference for AI-related image labels used by the Uranus public API and their relationship to EU transparency requirements.

Image AI labels describe whether and how artificial intelligence was used to create or modify an image.

Uranus provides the image_ai_label property in event responses to make this information available to websites and other API consumers.

Available labels

Value	Meaning	Recommended use
`none`	No AI generation or modification is declared.	The image is a conventional photograph or other image that has not been generated or modified using AI.
`ai`	AI was involved, but the specific type of use is not classified further.	Use when AI involvement is known, but it is not possible or appropriate to distinguish generation from modification.
`ai_generated`	The image was generated using AI.	Use for images created by an AI system, rather than captured as a conventional photograph or illustration created without AI generation.
`ai_modified`	An existing image was modified using AI.	Use when AI has been used to alter an existing image, for example by generating or replacing visual elements.

These values are Uranus reference labels. They are intended to communicate image provenance consistently across the API and its clients.

Choosing a label

Use the most specific label supported by the available information:

Choose none when no AI generation or modification is declared.
Choose ai_generated when the image itself was generated by AI.
Choose ai_modified when an existing image was modified using AI.
Choose ai when AI involvement is known but cannot be classified more specifically.

Do not assume that every use of an AI-powered editing tool necessarily requires the ai_modified label. Routine editing assistance may be treated differently under the applicable rules.

Relationship to the EU AI Act

Article 50 of the EU AI Act (Regulation (EU) 2024/1689) establishes transparency obligations for certain AI-generated and AI-manipulated content.

Depending on the circumstances, these obligations can include:

Machine-readable marking: Providers of AI systems that generate synthetic images and other content must ensure that their outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, subject to the conditions and exceptions in the Act.
Disclosure to people: Certain AI-generated or manipulated content, including content qualifying as a deepfake, must be disclosed to people exposed to it.
Appropriate presentation: Applicable disclosure must be clear and distinguishable. A machine-readable marker alone does not necessarily satisfy a separate obligation to inform people.

The four Uranus labels are an application-level classification. They do not, by themselves, establish compliance with the EU AI Act. A label stored in an API response is also not necessarily equivalent to embedding a machine-readable mark in the image file or providing a visible disclosure to users.

For the applicable scope, exceptions and practical guidance, consult the official European Commission resources below.

Transparency obligations under Article 50

Read the European Commission’s frequently asked questions about AI-generated content, marking, disclosure and exceptions.

Guidelines on AI transparency obligations

Consult the Commission’s guidelines for providers and deployers of AI systems.

EU AI Act — Article 50

Read the underlying regulation, including the transparency obligations for providers and deployers.

For API consumers

The image_ai_label property is optional in an event response and may be omitted when no value is available.

Example:

{
  "image_uuid": "01a09fed-c812-7d7e-b5ff-312de6185160",
  "image_path": "https://api.kulturbytes.de/api/image/01a09fed-c812-7d7e-b5ff-312de6185160",
  "image_ai_label": "ai_generated"
}

Clients can use this property to display a suitable disclosure or provide additional information about an image’s origin. The presentation should reflect the actual circumstances and any applicable legal requirements.