Since August 2, 2026, companies in the EU must openly disclose their use of AI, whether chatbot, AI image, or automatically generated text. This disclosure obligation, known as the transparency obligations of the EU AI Act, is now binding law for nearly all companies.
The EU AI Act, Europe's AI regulation, entered into force on August 1, 2024. Now, as of August 2, 2026, the core transparency obligations are being activated. This affects not only large corporations but almost all companies that use AI systems professionally. If you use chatbots in customer service, create AI-generated images for marketing campaigns, or produce automated texts, you need to act now.
The EU AI Act's transparency obligations have applied since August 2, 2026 to almost all companies that use AI professionally, requiring a visible interaction notice for chatbots and machine-readable labeling of AI-generated content. We will show you what exactly is required, who it affects, and how you can implement the requirements with reasonable effort. We hold a clear position: Companies that inventory their AI systems and build transparency processes now will avoid later adjustment pressure and liability risks.
Which transparency obligations apply from August 2026?
The regulation distinguishes between two central obligations. First: Providers of AI systems that interact with natural persons must ensure that these persons are informed that they are interacting with an AI. This applies to chatbots or AI-powered telephone assistants, for example. The information must be clear and understandable; a hidden footnote is not sufficient.
Second: Providers of AI systems that generate synthetic content must label this content as AI-generated in a machine-readable format. Synthetic content includes images, videos, audio, or text that has been entirely created by an AI. The labeling must be machine-readable, meaning a computer program can automatically detect it. This is typically done through metadata, i.e., additional information invisibly stored within the image or text file. This makes it possible to trace whether an image or text originates from an AI.
For deployers, meaning companies that use AI systems, this means concretely: If you publish AI-generated content on your website, you must ensure that this content is labeled accordingly. The obligation applies to you as the deployer, even if you did not develop the AI yourself. Our position: Labeling is not a burdensome formality but a foundation for trust and legally sound communication. Transparency builds credibility with customers and business partners.
Who is affected by the EU AI Act?
Since August 2, 2026, both AI providers and deployers are affected by the transparency obligations, each with different duties: Providers must technically build in notices and labeling, while deployers must ensure their effectiveness in use. The EU AI Act distinguishes between providers and deployers. Providers are the developers of AI systems who bring their products to market. Deployers are companies that use AI systems professionally under their own responsibility.
An example: A mid-sized company integrates an AI chatbot from an external provider into its website. The provider must technically enable the notice about AI interaction. The company, as the deployer, must ensure that the notice is actually displayed.
For existing AI systems, there is a possible relief: The not-yet-binding EU legal act 'Digital Omnibus on AI' plans a transition period until December 2, 2026. This act aims to reduce bureaucratic hurdles and postpones, among other things, the extensive obligations for so-called high-risk AI systems, meaning AI applications with particularly far-reaching impacts, such as in hiring decisions, credit scoring, or biometric identification, from August 2026 to December 2, 2027. Importantly: The Digital Omnibus is not yet in force. You cannot safely rely on it. We believe it would be a mistake to postpone preparation until the act becomes binding. The transparency obligations apply now, and preparation costs less time and money than reacting under time pressure later.
What does Article 4 require for AI literacy?
In addition to the transparency obligations, there is another requirement that is often overlooked: AI literacy under Article 4 of the EU AI Act. This requirement has been in effect since February 2, 2025. A planned, but not yet in force, revision of Art. 4 is part of the Digital Omnibus on AI. It requires that all persons working with AI systems possess sufficient knowledge to use these systems safely and responsibly.
For companies, this means: You must ensure that your employees understand what AI can do, where its limits lie, and what risks are associated with its use. This is not a one-time training but a continuous process. AI literacy is the foundation for complying with all other obligations. Only those who understand how an AI system works can assess whether transparency requirements are met.
We recommend viewing training not as a burdensome duty but as an opportunity. Employees who understand AI can, for example, correctly handle automatic content labeling and avoid errors when interacting with chatbots. The investment in AI literacy pays off twice: You meet legal requirements and simultaneously improve your workflows.
How do you implement labeling in three steps?
Implementation succeeds in three steps: 1) Inventory all AI systems in use, including hidden functions in CRM/image tools, 2) Assess each system for interaction or synthetic content with documentation, 3) Technically implement notices and labeling via metadata or watermarks. The first step is the inventory. Create a list of all AI systems used in your company. This includes not only obvious systems like chatbots but also AI functions in software you already use, such as CRM systems or image editing programs. Ask your departments which tools with AI functions are being used.
The second step is the assessment. Check each AI system to determine whether it interacts with natural persons or generates synthetic content. If so, you must establish whether the required transparency mechanisms are in place. For systems from external providers, clarify contractually who is responsible for labeling. Document your assessment; this creates evidence for future audits.
The third step is implementation. Ensure that interaction notices are displayed and content is labeled. Technically, this can be achieved through metadata, watermarks, or other machine-readable formats. Regularly check whether the mechanisms work and adjust your processes when AI systems change. A continuous improvement process is the right approach here.
Provider vs. Deployer: Who does what?
| Obligation | Provider | Deployer |
|---|---|---|
| Interaction notice | Technically build in | Ensure display |
| Labeling synthetic content | Implement metadata | Label publications |
Conclusion: Why acting now reduces costs
The transparency obligations of the EU AI Act are a reality. Companies that act now save time, money, and stress later. The inventory of AI systems is the foundation for everything else. It provides clarity about which systems are in use and what obligations arise from them. Those who postpone this work risk not only sanctions provided for in the AI Act but also a loss of trust among customers and business partners.
Our position is clear: Transparency is not an obstacle but a competitive advantage. Companies that communicate openly when AI is involved position themselves as trustworthy partners. Implementing the transparency obligations is achievable if you proceed systematically. Start with the inventory, assess your systems, and implement labeling. Your effort will pay off.