How Are AI Image Labels Changing? What Anyone Sharing Photos Should Know | Release #930

As generative AI evolves, looking at an image is no longer just about asking what it shows. Increasingly, we also want to know how it was made.

Questions around AI, including the authenticity of photos and photographers’ rights, are not new. What we need to understand now is that rules on labelling AI-generated content have entered the implementation stage.

The transparency obligations under Article 50 of the EU AI Act have applied since August 2, 2026. The European Commission’s code of practice sets out how providers and users of generative AI systems should handle the labelling and detection of AI-generated or manipulated content.

Providers are required to mark AI-generated or manipulated output in a machine-readable format and make it detectable. For users of generative AI, the obligations cover deepfakes that make people or events appear real, as well as certain AI-generated or manipulated text on matters of public interest. What is required varies depending on the content and whether someone is providing or using the system.

Taking part in the code of practice is voluntary. However, the European Commission makes clear that the transparency requirements under Article 50 remain legal obligations even for those who do not participate. The code provides a practical framework for meeting those requirements.

This means it would be misleading to say that every image involving AI needs the same label, or that every photo adjusted after shooting should be treated as an AI image. The obligations that apply depend on factors such as whether you provide or use a generative AI system and what kind of content you create and publish.

Start by keeping your original photo files separate from any generated or manipulated versions. If you use generative AI, keep a record of where it was used in your process. It is also worth checking the labelling tools and guidelines of the platform where you post, so you can explain your process later if needed.

When sharing an image that appears to show a real person or event, add enough context to prevent viewers from being misled. This does not mean you need to detail every edit. Trust comes from accurately explaining how an image was made when that information matters.

It is becoming harder to tell from appearance alone whether an image is a photograph or was generated by AI. That is why we need to consider not only what an image shows, but also how it was made and how it is described when published.

It is becoming harder to tell from appearance alone whether an image is a photograph or was generated by AI. That is why we need to consider not only what an image shows, but also how it was made and how it is described when published.