For companies that use AI systems, whether for marketing, content creation, customer service or data analysis, it is crucial to keep a close eye on these developments. , who it applies to, and how you can prepare for it.
What is the code of practice for AI transparency?
The code of practice is a voluntary guideline for marking and labelling AI-generated or manipulated content. This includes, for example, texts, images, audio or video created using generative AI systems. The code of practice was drawn up by independent experts, with input from hundreds of stakeholders from the business community, academia and civil society organisations.
The code focuses on two groups:
- Providers of generative AI systems: for example, companies that develop or offer AI models.
- Operators or users of these systems: such as companies that use AI to generate content, for example for websites, social media or internal communication.
What are the core obligations?
The AI Regulation requires providers and users of generative AI systems to be transparent about the use of AI. This means:
- Labelling: Providers must ensure that AI output (such as generated text or images) is labelled in a machine-readable format. This can be done, for example, using metadata, digital watermarks or fingerprinting techniques.
- Detection: The content must also be detectable as AI-generated, so that platforms or other parties can recognise that it is synthetic content.
- Labelling: Users (operators) of AI systems must make it clear when the content they publish is artificial. This applies in particular to so-called deepfakes or texts intended to inform the public about social or political issues.
An example: deepfake video
Suppose your company uses an AI tool to create a video in which a virtual speaker explains a product. If that video closely resembles an existing person, it may be considered a deepfake. In that case, you must clearly indicate that the video is artificial – for example, with a label or icon during playback.
An example: AI-generated news articles
A company that uses AI to automatically publish news reports on economic trends must state that the text has been generated by AI. Unless a human editor checks the text beforehand and bears final responsibility.
What does this mean in concrete terms for entrepreneurs?
Although the code of practice is currently voluntary, in practice it serves as the standard by which regulators and consumers will assess whether an organisation is dealing with AI in a ‘transparent’ manner.
For entrepreneurs, this means:
- You need to know whether the tools you use comply with labelling requirements.
- You must develop policies on when and how you label AI content.
- You should consider whether human control is necessary for publications that have an impact on public debate.
- You should be prepared for questions from customers or regulators about how your company uses AI.
What should you do now?
- Map out your AI usage: Which tools do you use? Do they generate content? Is this content shared publicly?
- Check whether your suppliers comply with Article 50: Do they meet the technical requirements for marking and detection?
- Consider self-labelling: Determine whether you need to actively label certain content, for example on your website or social media.
- Document your approach: Keep internal records of how you handle AI content, which tools you use, and what oversight is in place.
- Provide training and awareness: Employees who use AI need to know what the rules are and how to apply them. AI trainer and solicitor Julia van Leeuwen can provide this training for your company or department.
Why these rules?
The European Commission wants to combat deception, manipulation and fraud involving AI content. Deepfakes in particular carry a high risk of reputational damage and confusion. Think of manipulated videos in which politicians or CEOs appear to make statements they never actually made.
At the same time, the EU does not want to hinder innovation. That is why a code of practice has been chosen as a tool to help companies deal with AI responsibly.
What else is coming?
The current code of practice is still in draft form. Stakeholders can provide feedback until 23 January 2026. The final version will follow in June 2026. The transparency obligations will become legally binding from August 2026.
In addition, the Commission will also publish guidelines to assist in interpreting the rules. These will include examples, exceptions and sectoral nuances.
Conclusion
AI offers entrepreneurs many advantages, but also requires caution. Transparency is essential in this regard. By already working on marking and labelling AI content, you can avoid legal risks and strengthen the trust of customers, partners and the public.
Do you have questions about how these rules affect your business? Would you like assistance in drafting AI policies, transparency statements or terms and conditions? Please feel free to contact Julia van Leeuwen at Guldemond.
