Press release: Transparency obligations are essential elements of the AI Act
The political compromise reached between the EU Parliament, Council and Commission on the EU AI Act on 8 December is now being finetuned at technical level. The SAA welcomes the agreed upon transparency obligations of general-purpose AI models’ providers operating in the EU market. The technical discussions must now consolidate the political agreement, without weakening its scope, and make transparency a reality.
Transparency at both the input and output stage is necessary for AI to be developed and deployed responsibly, and for rightsholders and their collective management organisations, to be informed when their works are being used.
Obligations for transparency at the input level ensure quality data being ingested in AI models. They provide a clear and necessary framework for both developers and creators. We welcome the obligation of providers of general-purpose AI models to draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model and we are ready to work with the future AI Office on a template to would address the needs of the authors. Transparency is also needed at the output level and we call for clear rules on the labelling of AI generated production.
We also welcome the obligation for all general-purpose AI models’ providers operating in the EU market to respect EU copyright law. While the AI Act should not interfere with the substantive copyright rules which are an autonomous corpus, an obligation to put in place a policy to respect Union copyright law is a good reminder for developers to protect and promote the continued development of human creativity and original works.
Quote
“The EU AI Act is the first regulation of its kind in the world and a one-time opportunity for EU Member States to be leaders of innovation that protects and promotes the continued development of human creativity and original works. The devil is in the details, so we trust the technical negotiators to be vigilant to ensure the spirit of a human-centric and trustworthy legislation.”
Barbara Hayes, Chair of the Society of Audiovisual Authors
Background
The SAA published a comprehensive position paper in October 2023 analysing the use of audiovisual works as training data for machine learning, the status of AI generated audiovisual production and developed principles for a human-centred AI regulation that fosters creativity. Such crucial principles are authorisation, remuneration and transparency.