AI Act Transparency Obligations - What They Mean in Practice?

The EU AI Act requires organizations to clearly disclose AI use. Whenever a customer is interacting with an AI chatbot or a marketing image representing a real world object is made with AI, AI use should be disclosed. This is an obligation that many organizations have not yet addressed.

Why transparency obligations exist?

AI systems used for client communication, text and image generation and customer service are becoming increasingly common. As generative and interactive AI systems become part of everyday business tools, it is getting harder for people to tell the difference between AI-generated content and something a human made.  

People should know when they are interacting with AI or exposed to AI-generated content. That is why the EU AI Act includes transparency requirements to protect the public from fraud, manipulation, impersonation and consumer deception.

What this means for providers?

Article 50 of the AI Act sets out transparency obligations for providers and deployers of certain AI systems, including generative and interactive AI systems and deepfakes. 

If your organization builds or supplies an AI system that falls in scope, you must:

• Design the system so people are explicitly informed whenever they are interacting with an AI system directly (for example, a customer-facing chatbot).
• Add machine-readable marks to AI-generated or AI-manipulated content, so that it can be detected as such. 

What this means for deployers?

If your organization uses AI systems built by someone else, your obligations are on the receiving end of that interaction. You must inform individuals when they are exposed to:

• Emotion recognition or biometric categorisation tools
• Deepfakes
• AI-generated text published on matters of public interest, where there has been no meaningful human review or editorial control.

Practical recommendations to avoid enforcement

Compliance will be monitored by national market surveillance authorities and the EU AI Office. A few practical recommendations to get started with Article 50 obligations.

Inventory your AI touchpoints. Map how your organization uses AI chatbots, generative content creation tools and emotion recognition/biometric recognition systems. Determine if these systems fall under article 50 obligations. 

Design the disclosure. Design how you will fulfil the informing obligations - is it a watermark, banner within the chatbot window or a clear on-screen notice at the point of exposure? Consider adherence with a EU Code of Practice or documenting your own equivalent measures.

Build evidence. Document how you have fulfilled the transparency obligations and what measures you have put in place. Logging AI transparency trainings and workshops for additional evidence is a great additional step.

References and additional reading:

Code of Practice on Transparency of AI-generated Content

EU AI Act: Article 50 Transparency obligations for providers and deployers of certain AI systems