Statement: Photo Council on AI

The associations and institutions organized in the German Photographic Council take a position on AI image generators in a joint statement. Member of the team of authors was ProfiFoto editor-in-chief Thomas Gerwers in his capacity as chairman of the DGPh section Art, Market and Law.

Here is the wording of the statement:

Methods for processing and generating images with the help of artificial intelligence (AI) will have a fundamental and disruptive influence on work processes in all areas of photography. The German Photographic Council is fundamentally open to the further development of imaging techniques. Generative AI opens up new possibilities for expanding the creative spectrum. However, it will also bring serious economic upheaval to the creative industry and may pose a serious threat to democratic social discourse.
The Photo Council calls on all institutions involved in the creation and dissemination of news documentary content to develop ethical standards for dealing with their sources. These rules and working practices must ensure that authentic material is verified and passed on as such in a recognizable and verifiable manner. In clear distinction, generated images should not be called photographs, even if their photorealism is reaching ever higher levels.
Without the creative act of creation by humans through the inclusion of new pictorial works and the formulation of original prompts, AI image generators tempt people to reproduce and vary what already exists, since they always generate their images by recourse to already existing content. However, central legal issues when working with AI image generators are currently unresolved. For example, it is disputed under what circumstances photographers can obtain copyright to their works when using AI tools and what rights users of these works are entitled to.
Without permission and compensation, AI systems use photographers' intellectual property as training material or disguise the origin of data. The Photo Council calls for more speed in legal clarification to create legal certainty in the use of AI tools and a basis for the further development of creative professionals' business models.

AI images are not photographs
It is important to the German Photo Council to distinguish between camera-based photographs and synthetically generated AI images, because photographs are created exclusively by imaging light in a camera. AI-generated images are therefore not photographs and should not be referred to as such, even if their photo-realistic presentation gives them this impression. Because the qualitative boundaries disappear, a clear differentiation between photographs on the one hand and generated images on the other is crucial for their classification and perception and strengthens the media literacy of viewers.
This also touches on aspects of photographic pedagogy, the goal of which must be a "school of seeing" that guides even children and young people to also critically question images. With regard to democratic opinion-forming, there is an urgent need to raise awareness of the importance of consulting multiple sources or eyewitnesses.

Consequences for authors
The unsolicited exploitation of their copyrighted image material by means of data mining for generally accessible image generators means an economic loss for photographers that threatens their existence. This is because the current mode of operation of AI systems is contrary to the fundamental principle of copyright law that authors themselves and alone reap the rewards from the exploitation of their works and receive appropriate remuneration.
The German Photo Council would like to see the fastest possible legal certainty with regard to the copyright and exploitation rights of the creators of photographs used as training data. It must be traceable on what basis an AI image was generated. If images from photographers are processed for this purpose, there must also be mechanisms for fair compensation for creators of these images.

Transparency in AI training material
In any case, creators must be able to exercise the right of disposal over the use of their works and object to the use of their images by AI systems. Photographers currently have only very limited possibilities to do this. Therefore, the Photographic Council calls for simple opt-out or opt-in options. For example, photographers can opt-out of the use of their works as AI training material or opt-in to explicitly release their works for this purpose as soon as practicable remuneration mechanisms for this use have been established. Initial approaches are the "Do not train" certificates of the Content Authenticity Initiative (CAI). The German Photographic Council has become a member of the CAI.

In particular, the Photographic Council demands that the legislator immediately define how the legally provided reservation against the use of image material for the purpose of so-called data mining can be made concretely in "machine-readable form". For this purpose, the Fotorat strives for a regulation in which the embedding of such a reference in the EXIF/IPTC data of an image file is sufficient.
At the same time, the Photographic Council agrees with the demand that such metadata must not be separated from or deleted from the associated image data, as is often done today when uploading images to platforms or when using images.
The German Photographic Council supports the proposal to review in the short term the non-profit status of the datamining organization Large-scale Artificial Intelligence Open Network (LAION) registered in Germany, which makes works available on a large scale to AI system providers for commercial use as training material without the consent and appropriate remuneration of the creators.

Ensure authenticity
Authentic photographs are essential for the credibility of images in journalistic media. Distinguishing between generated and camera-based images is already difficult or impossible, even for experts. Democratic discourse is endangered by manipulation by means of convincing image forgeries, but just as much by the general doubt about the authenticity of images, which will be justified in the future.
Photographers must take responsibility for what they photograph, how they photograph and edit it, and to whom they pass on their images.
The Photographic Council therefore supports the demand for the development of internationally uniform and consistently open technical standards for the verification of image creation in cameras and for the logging of subsequent processing steps.
To ensure their credibility, the Photographic Council calls on image users in the media in particular to mark image material clearly visible similar to the copyright notice directly on the image according to its origin.

Copyright in AI images
Copyright is a protection right for human intellectual creations. The Photographic Council suggests that the recognition of copyright protection of AI-generated images be examined on the basis of existing law. Solutions need to be worked out as to how to draw the line between human computer-assisted design that can be protected by copyright and presumably purely computer-generated machine product in which, according to widespread legal opinion, no copyright can be acquired. Given the disruptive nature of AI tools, it is also necessary to question whether the classic tools for assessing copyright need to be adapted.
We urgently and quickly need legal certainty both for the creators of AI-generated images and for users and clients of such works.

Conclusion
The Photographic Council understands AI systems as an additional, new possibility for image creation, whose products, however, must be clearly distinguished from photographs. It urges regulations as soon as possible to prevent AI systems from being trained and applied to the detriment of creators of camera-based photographic images.
The impact of the new systems on our society is already becoming clear. By combining synthetic images, which can be disseminated in almost any quantity, with AI-generated text and sound documents, a complex web of seemingly coherent pseudo-realities can be created.
Especially against the backdrop of working conditions in most areas of the media industry that are geared towards economic optimization, serious efforts to effectively counter the danger posed by manipulative content created with AI tools are hardly discernible at present. Declarations of intent are not enough for this. Protecting democratic discourse requires binding editorial guidelines and adequate resources to implement them.

https://deutscher-fotorat.de