From this Sunday, 2 August, every company placing an artificial-intelligence system on the EU market must comply with the first binding transparency and labelling duties of the bloc's AI Act, according to Demócrata and Tech Review Africa. The rules do not ban any technology. They require providers and deployers to disclose, in specific and legally defined ways, when a user is talking to a machine, when an image or video is synthetic, and when a published text was AI-written — and the duty applies wherever the company selling into the EU happens to be based.

The legal mechanics: Article 50

The duties sit in Article 50 of Regulation (EU) 2024/1689, the AI Act, which entered into force in August 2024 and is being phased in over three years, per the European Commission's own transparency-rules guidance and analysis by Stibbe. Article 50 covers four separate situations: systems that interact directly with people, such as chatbots, must make that clear unless it is obvious; AI-generated or manipulated content, including deepfakes, must carry a label that is detectable and stays attached when shared, per reporting by The Next Web; emotion-recognition and biometric-categorisation systems must inform the people exposed to them; and AI-generated text on matters of public interest must be flagged unless it underwent human editorial review, per the Article 50 guide.

Breach of Article 50 carries fines of up to €15 million or 3 percent of worldwide annual turnover, whichever is higher, according to the Commission's own factsheet. That is a lower, second-tier band: the Act's headline penalty, up to €35 million or 7 percent of global turnover, is reserved for the short list of outright prohibited practices under Article 5 — social scoring, manipulative systems, certain biometric categorisation — which has applied since 2 February 2025, per MediaLaws' compliance tracker. It is Sunday's lower band, not the 7 percent maximum, that governs the transparency duties taking effect this weekend.

Who enforces, and how

Like the GDPR, the AI Act polices itself through national bodies coordinated from Brussels. Day-to-day supervision falls to a "market surveillance authority" designated by each of the 27 member states, the same architecture the EU uses for product-safety law generally, per Bird & Bird's analysis of the Commission's final Article 50 guidelines. Above the national authorities sits the EU AI Office, which supervises general-purpose AI models directly and coordinates enforcement across the bloc; where an EU institution is itself the provider or deployer, the European Data Protection Supervisor takes that role instead.

Hypothesis: enforcement is likely to be uneven at first, echoing the slow, fragmented start the GDPR had under national data-protection authorities after 2018. Supporting this: trackers monitoring the rollout, including Stibbe and Bird & Bird, note that the Commission's Article 50 guidelines were only finalised shortly before the deadline, leaving national regulators little time to build enforcement capacity. Against this: the AI Office has already exercised comparable authority for a year, since its powers over general-purpose AI providers took effect on 2 August 2025, per MediaLaws — so the institutional machinery is not starting from zero.

What this means if you...

  • Run a customer-facing chatbot or AI voice service sold to EU users: you generally need a clear, upfront disclosure that the user is interacting with AI, unless that is self-evident from context.
  • Publish AI-generated images, audio, video or text: content must carry a label that survives sharing and, for deepfakes, is machine-detectable; a grace period for marking generative systems already on the market before 2 August 2026 runs to December 2026, per Analytics Insight.
  • Operate emotion-recognition or biometric-categorisation tools in the EU, even for internal HR or security use: you must inform the people being analysed, separately from any consent obtained under the GDPR.

Timeline and open questions

Sunday is the third of five phase-in dates. Prohibited practices and AI-literacy duties took effect 2 February 2025; the general-purpose-AI governance chapter and the penalty framework itself began 2 August 2025; this weekend brings transparency duties and most high-risk-system obligations; a final tranche, covering AI embedded as a safety component in already-regulated products such as medical devices or machinery, follows on 2 August 2027, per the Implementation Timeline tracker. Open questions include how strictly national authorities will interpret "detectable" labelling for deepfakes in practice, and whether the Commission's voluntary Code of Practice on AI-generated content hardens into a de facto compliance floor, as its general-purpose-AI counterpart already has.

This article is general information, not legal advice. Whether a specific product or service falls within Article 50, and which obligations apply, generally depends on the circumstances and should be assessed case by case.