Why the AI transparency rules affect more businesses than many think
On 10 June 2026, the European Commission published its Code of Practice on Transparency of AI-Generated Content, the final piece of guidance before the transparency obligations in Article 50 of Regulation (EU) 2024/1689 (the AI Act) began to apply on 2 August 2026. After the Digital Omnibus postponed the high-risk requirements, many companies concluded that the whole regulation had been put on hold. That is a misunderstanding. The transparency rules were never postponed. They have applied since 2 August 2026 and are enforceable today. One narrow transitional period remains: generative systems already on the market before that date have until 2 December 2026 to add machine-readable marking of synthetic content under Article 50(2).
Unlike the high-risk requirements, which target a limited group of systems, the transparency obligations reach almost any company that uses AI in customer contact or content production. A customer service chatbot, an AI-generated campaign image or an AI-written article can all trigger obligations.
What is Article 50 of the AI Act?
Article 50 contains four separate transparency obligations, each aimed at a different situation. Some apply to companies that develop and supply AI systems, others to companies that simply use the technology in their business. The common idea is that people should know when they are talking to a machine or looking at machine-made content. Non-compliance can result in fines of up to 15 million euros or three percent of global annual turnover, whichever is higher.
Which obligations apply to your business?
Chatbots and other interactive AI
Companies that provide AI systems intended to interact directly with people, such as customer service chatbots or voice assistants, must design them so users understand they are dealing with AI. The obligation does not apply where this is obvious to a reasonably well-informed person, but the Commission’s draft guidelines from May 2026 interpret that exception strictly. A clear notice at the start of the conversation is the safest solution.
Marking of AI-generated content
Companies that provide generative AI systems must ensure that synthetic audio, images, video and text are marked in a machine-readable format so the content can be detected as artificially generated. Under the Digital Omnibus, systems already on the market before 2 August 2026 get a transitional period until 2 December 2026 to meet the machine-readable marking requirement.
Emotion recognition and biometric categorisation
Companies that use systems for emotion recognition or biometric categorisation must inform the people exposed to them. Bear in mind that emotion recognition in workplaces and education is already prohibited under Article 5 of the regulation, so this obligation mainly becomes relevant in areas such as customer analytics.
Deepfakes and AI-written text
Companies that use AI to generate or manipulate images, audio or video that realistically resemble real people, places or events, known as deepfakes, must disclose that the content is artificial. According to the draft guidelines, content that is clearly unrealistic, such as fantasy creatures, falls outside the definition, and minor technical edits do not make material a deepfake. AI-generated text published to inform the public on matters of public interest must also be disclosed, unless it has undergone human review and someone holds editorial responsibility.
What counts as sufficient disclosure?
Information must be clear, distinguishable and provided no later than the first interaction or exposure, and it must meet applicable accessibility requirements. A statement buried in terms and conditions does not satisfy the requirement, according to the draft guidelines. The new Code of Practice is voluntary but offers practical help, including uniform EU icons and text labels for deepfakes and guidance on where to place them. Following it is an easy way to demonstrate a serious approach to compliance.
A practical example: an online retailer
A Swedish online retailer uses a chatbot for customer service, produces campaign images with generative AI showing realistic models in real environments, and drafts product descriptions with an AI assistant. From 2 August 2026 the chatbot must announce that it is AI at the start of each conversation. The campaign images are deepfake-type content and must carry a visible disclosure, and the image generator used must embed machine-readable marking. The product descriptions, by contrast, are not aimed at informing the public on matters of public interest, so they require no disclosure, provided the retailer reviews them before publication.
Common mistakes companies make
The most common mistake right now is believing that the Digital Omnibus postponed everything, and pausing transparency work that must be finished by August. Another is burying the AI notice in terms and conditions or a privacy policy, which the Commission has explicitly said is insufficient. Many companies also assume the rules only concern AI developers, when several obligations fall on any business that deploys the technology. Finally, some read the transitional period until December 2026 as a general delay, when it only covers machine-readable marking for systems already on the market.
Recommended actions
Begin by mapping every AI system your business provides or uses in customer contact and content production, and note for each whether you develop it or merely use it. Update chatbots and voice assistants so they announce themselves at the first interaction. Review how marketing and communication teams use generative tools, and introduce routines for labelling realistic AI-generated material. Check that the generative systems you rely on embed machine-readable marking, and ask your suppliers for confirmation. Use the Code of Practice icons and labels rather than inventing your own, follow the final guidelines expected before August, and document your decisions as you go.
Frequently asked questions
Do the transparency rules apply to small businesses?
Yes. Article 50 contains no size threshold, so a small company with a chatbot or AI-generated marketing content is covered in the same way as a large one. The practical effort is usually modest once systems and content flows have been mapped.
Were the transparency rules postponed by the Digital Omnibus?
No. The Digital Omnibus postponed the high-risk requirements but left Article 50 on its original schedule. The only relief is a transitional period until 2 December 2026 for machine-readable marking in generative systems placed on the market before 2 August 2026.
What happens if we do not comply?
Breaches of the transparency obligations can lead to fines of up to 15 million euros or three percent of global annual turnover, whichever is higher. National supervision is still being organised in several member states, but the obligations bind companies from 2 August 2026 regardless.
Conclusion
The transparency rules are the part of the AI Act that will touch most businesses first, and they take effect on 2 August 2026. Companies that use the coming weeks to map their AI use, adjust chatbots and set up labelling routines can meet the deadline with limited effort, while those who wait risk fines and hasty fixes. The Commission’s guidelines and the new Code of Practice make compliance easier, not harder. At Lawgent, we help companies translate the transparency obligations into concrete measures, from mapping and role analysis to labelling routines and supplier requirements. Get in touch if you want to know what Article 50 means for your business before August.
