The transparency obligations introduced by Article 50 of the EU AI Act (Act) apply to AI providers (organisations that create or supply AI systems) and deployers (organisations that operate such systems as part of their business activities). The Act’s requirements are being phased in gradually and the transparency obligations that came into force on 2 August are one of the significant implementation milestones. The purpose of the Article 50 transparency obligations is to help individuals understand whether content is AI-generated.
B P Collins’ corporate and commercial team explores the requirements under the Act, which can extend beyond the EU to the UK and other territories where AI systems and resulting AI-generated material are used within the EU.
What are the new regulations?
Overall, users must be clearly informed when they first encounter/interact with the AI system or its output (i.e., AI systems that generate or edit videos, images, audio and text content). Providers must disclose direct AI interactions and mark AI-generated or altered content as AI-generated or manipulated. There are four key quality requirements for solutions that detect and mark AI system outputs:
- Effective – reliably distinguishing AI content from human content.
- Reliable – accuracy of identifying and distinguishing AI content.
- Robust – withstanding attempts to remove or alter markings.
- Interoperable – allowing integration with other networks and solutions.
Providers of systems that are now placed on the market since 2 August 2026 must comply with the transparency requirements straight away. Providers of systems that were already on the market before 2 August 2026 have until 2 December 2026 to comply with the obligations.
There are exceptions to the transparency obligations, for example, for AI systems carrying out standard editing or where AI systems do not substantially alter the original input data.
Deployers are obliged to label content that are deepfakes, but can be done in such a way that does not hamper the display or enjoyment of the work where the content forms part of an artistic, creative or fictional piece of work. Also, if deployers generate or manipulate text which is to be use for informing the public of public interest matters, they must also label such AI-generated or manipulated text (except where the use is legally authorised for law enforcement purposes, or where the AI-generated content has been subject to human review or editorial oversight).
The EU Commission has published Guidelines on transparency obligations for providers and deployers of AI systems which include examples of how a transparency notification can be provided to users (such as a chatbot starting a conversation stating that it is based on AI technology, or a voice assistant stating at the start of a session that it is powered by AI technology). The EU Commission has also published a Code of Practice on Transparency of AI-generated Content, which provides a voluntary framework to demonstrate compliance.
Organisations should be aware that non-compliance with the Act’s transparency obligations could result in a financial penalty of up to €15m or 3% of global turnover over the previous financial year – whichever is higher.
For further advice on how the EU AI Act could affect your business, including if you are using customer-facing chatbots or AI assistants, or generating AI-created content, please contact B P Collins’ corporate and commercial team at enquiries@bpcollins.co.uk or call 01753 889995.
