AI image generation technology opens up exciting creative possibilities for our customers, but it must be approached in the right way to protect the interests of Alamy content creators. As part of our commitment to exploring ethical opportunities within AI we are pleased to announce our participation in the ‘The Fair Diffusion Program’ – a ground-breaking generative AI initiative launched by BRIA AI in collaboration with Alamy, Getty Images, and Envato. This is an alternative to the AI models already in existence that have been built on unlicensed content – instead it’s designed to attribute and reward source material.
Compensation, Recognition, and Transparency
Central to the ethos of The Fair Diffusion Program is a commitment to fairly compensate photographers, artists and creators for the source material used to generate each AI image. The AI model built by Bria has been trained on licensed content from Alamy, Getty Images, Envato and other partners, but what makes it truly unique is its attribution model. Each time an image is generated through the model, reverse technology will identify the source material used to arrive at that individual generated image. This means that contributors can receive commission for each generated image that is sold.
We know that AI image generation is currently controversial and that its legal framework has yet to catch-up with the fast pace of development and adoption. Most of the major models in existence do not recognise source creators, and make no attempt to recompense or credit them. While it is early days for the program, we believe an ethical alternative to other models in existence is the right approach.
Training for this new model is continuing and users are testing. Once the generated images start being used commercially the attribution framework will kick in and Alamy contributors will receive their share. We will keep you updated as this begins to take shape and will be transparent about how the mechanism will work.
The collection used by Bria to train the Fair Diffusion model is ‘Novel Use’ and covered by our terms and conditions. A similar collection was submitted by Getty Images and Envato. Being part of ‘Novel Use’ at Alamy is all about being open to experimental new use cases for images and this includes AI.
Our inclusion in the programme will also help to shape an AI model with greater diversity because of the open and inclusive nature of the Alamy platform.
https://bria.ai/
What’s next for AI and machine learning?
As well as The Fair Diffusion Program there are a number of other developments and steps being taken around AI and machine learning at Alamy:
1. We continue our involvement as part of CEPIC and through the DMLA and BAPLA to lobby for the introduction of the draft EU AI Act and also for recognition and fairness for copyright holders under pending UK guidelines.
2. We are exploring machine learning licensing opportunities for generative AI where we can transparently recompense our contributors for providing the source material.
3. We continue to license content for AI outside of image generation – such as image recognition, voice assistance and healthcare.
4. We are open to receiving commissions for specific image content to train specific AI models and will make these opportunities available to our suitable contributors directly.
5. While the sources for many AI image generation tools are at worst scraped or at best unclear we cannot accept AI generated material on Alamy and we continue to find and remove this. When The Fair Diffusion Program allows an alternative, we may explore a dedicated AI collection on our platform that is clearly labelled (and where there is a clear way to pay commission to those creators who are the originators of the source material).
6. Alamy will continue to discuss this changing landscape with our partners, friends, contributors and competitors and keep you informed.