News
3 min read
Companies operating AI chatbots in the EU or distributing AI-generated content have been required since 2 August 2026 to disclose this to users. The transparency obligations under Article 50 of the EU AI Act entered into force on that date, two years after the broader regulation took effect on 1 August 2024.
The rules address two specific situations: the use of AI in conversational systems, such as chatbots, and the distribution of synthetic media, including deepfakes. Providers and deployers of such systems must actively inform users that they are interacting with an AI system, or that the images, audio or video clips they are seeing have been artificially generated or manipulated.
For non-compliance, supervisory authorities may impose fines of up to €15 million or 3 percent of global annual turnover, whichever is higher. For EU institutions themselves, the maximum is €750,000. Enforcement for small and medium-sized enterprises will take proportionality into account.
Article 50 of the AI Act places obligations on both providers and deployers of AI systems. Providers of chatbots and similar conversational systems must build in the technical capability to inform users that they are communicating with AI. Deployers, the parties that actually make such systems available to end users, are responsible for actually displaying that notification.
A separate obligation applies to synthetic media. Anyone distributing AI-generated or AI-edited images, videos or audio recordings must indicate this in a clearly recognisable manner. The rules are specifically intended to make deepfakes visible to the public, both in commercial and political contexts.
There are exceptions. Where the use of AI is self-evident, for example in an animation clearly presented as fictional, or where content is created solely for personal use and not further distributed, the disclosure requirements do not apply. Some latitude also exists for legitimate purposes such as satire or parody, provided this is clear from the context.
Alongside the statutory obligations, the AI Office in Brussels has facilitated the development of a Code of Practice on transparency for AI-generated content. This document was drawn up by independent experts in a process involving multiple stakeholders. By the end of July 2026, approximately 190 companies and organisations had signed the code.
The Code of Practice does not replace the law, but offers companies a practical framework for meeting the obligations. Signatories commit to concrete measures in the areas of labelling and detection of AI-generated content. The AI Office also facilitates enforcement of the broader AI Act, together with national market surveillance authorities. For EU institutions themselves, the European Data Protection Supervisor has been designated as the supervisory authority.
Not all parties are unreservedly positive about the guidelines. The Computer and Communications Industry Association, a lobby group representing major technology companies, and the retail lobby group Eurocommerce have raised objections. Their concerns relate, among other things, to the feasibility of the obligations and the lack of clarity about what exactly qualifies as AI-generated content in complex production chains, where human and automated processing are intertwined.
Henna Virkkunen, Executive Vice-President of the European Commission for Technological Sovereignty, Security and Democracy, has defended the rules as necessary for citizens' trust in digital environments. Sergey Lagodinsky, a Green Member of the European Parliament who was involved in the AI Act negotiations, previously emphasised the importance of transparency for democratic processes, particularly regarding synthetic media in campaign communications.
The EU AI Act, formally Regulation (EU) 2024/1689, entered into force on 1 August 2024. The legislation follows a phased implementation: the strictest rules for so-called prohibited AI applications have applied since February 2025, while the transparency provisions have now come into effect as the second major phase. Rules for providers of general-purpose AI models followed earlier in 2025. The full regulation, including the obligations for high-risk AI systems, will become fully applicable in 2027.
For Dutch and other European companies using AI in customer contact or content production, 2 August marks a concrete compliance milestone. Responsibility lies with both the providers of AI technology and the organisations deploying that technology. This distinction is relatively new for many companies: a business that procures an off-the-shelf AI chatbot service and makes it available to customers bears, as a deployer, an independent legal obligation, separate from whatever the provider of the underlying technology has arranged. For investors and founders in the Dutch AI sector, this means compliance is now a structural component of product development and business operations, rather than something addressed only at market launch.
This article draws in part on the following sources.
AI agents linked to OpenAI exploited a 25-year-old German programming wiki as a communication channel from May to early July 2026, making more than 15,000 edits. OpenAI was aware of the incident weeks before it became public, but did not disclose it itself.
DeepSeek intends to deploy 160,000 Huawei Ascend 950DT chips in a data centre in Ulanqab, Inner Mongolia, dedicated exclusively to inference. Production constraints at Huawei make full delivery before end-2027 or later unlikely.
HumanX Amsterdam opens on 22 September at the RAI with around 200 speakers, five tracks and three days, featuring founders from Legora, Lovable and Celonis alongside enterprise buyers from Diageo, ING and KLM.
