From the journal

EU AI Act Article 50 Transparency Rules Become Enforceable on 2 August 2026

From 2 August 2026, Article 50 of Regulation (EU) 2024/1689 (the EU AI Act) is enforceable across all Member States. The Article imposes disclosure obligations on providers and deployers of interactive AI systems, emotion recognition tools, biometric categorisation systems, and generative AI producing synthetic audio, images, video, or text. The European Commission published accompanying guidelines on 30 July 2026.

3 min read

Article 50 of Regulation (EU) 2024/1689, the EU AI Act, became enforceable on 2 August 2026, applying to providers and deployers of AI systems across all EU Member States. The obligations are at the final-rule stage; they do not require further implementing measures. Article 50 is not limited to high-risk AI systems: it applies to any AI system falling within four defined categories regardless of risk classification.

The Article establishes four duties. First, providers of AI systems that interact directly with natural persons must inform those persons that they are interacting with an AI system, pursuant to Article 50(1) of Regulation (EU) 2024/1689. Second, providers of emotion recognition systems and biometric categorisation systems must inform the natural persons exposed to those systems, pursuant to Article 50(3). Third, deployers of AI systems generating deepfakes on matters of public interest must disclose the AI origin of the content, pursuant to Article 50(4). Fourth, providers of AI systems generating synthetic content must ensure that outputs carry machine-readable markings detectable as AI-generated, pursuant to Article 50(2); the AI Omnibus provisional agreement of May 2026 defers this machine-readable marking obligation for systems already on the market before 2 August 2026 until 2 December 2026.

The obligations fall on two classes of operators. Providers of interactive generative AI systems, including builders of chatbots, writing assistants, and synthetic media tools built on general-purpose AI models, must implement machine-readable marking controls and user disclosure mechanisms before deploying their products in the EU. Deployers of emotion recognition or biometric categorisation systems must notify employees, customers, or other persons exposed to those systems in advance of each use. Providers and deployers that fail to comply face administrative fines imposed by national market surveillance authorities and, for general-purpose AI model providers, by the European AI Office.

The deferral applies only to the machine-readable marking obligation under Article 50(2) for systems already placed on the market before 2 August 2026; no equivalent extension covers the human-facing disclosure duties under Articles 50(1), 50(3), or 50(4). The AI Office retains supervision of general-purpose AI model providers; national authorities supervise other deployers. The European Commission's guidelines of 30 July 2026 clarify scope and provide practical examples for each category of obligation.

Licentium advises technology providers and enterprise deployers on compliance with EU AI Act requirements. Work we undertake includes Article 50 disclosure obligation assessments, machine-readable marking implementation reviews, general-purpose AI model governance audits, and regulatory liaison with national AI authorities and the European AI Office.

Source: European Commission, Guidelines on Transparency Obligations for Providers and Deployers of Certain AI Systems under Article 50 of Regulation (EU) 2024/1689, 30 July 2026

More from the journal

See all
Illia Prokopiev

MLR Registration and the FCA Cryptoasset Gateway to 25 October 2027

This matter concerns the transition of a United Kingdom cryptoasset business from FCA registration under the Money Laundering Regulations 2017 to Part 4A permission under the Financial Services and Markets Act 2000. The question is whether MLR registration gives conversion, grandfathering, priority, or a right to continue after 25 October 2027, and what an affected firm should do before the gateway closes. This analysis assumes an existing UK-facing cryptoasset business, no relevant Part 4A permission, and an intention to continue after commencement.

Illia Prokopiev

ESMA's 2026 Custody Resilience CSA and the Rules That Actually Bind

ESMA’s 2026 Common Supervisory Action is a coordinated national review of digital operational resilience in crypto-asset custody. It will test whether selected crypto-asset service providers can demonstrate effective controls across six announced workstreams. The legal questions are which requirements are binding, how national competent authorities may assess control effectiveness, and what consequences may follow from a deficiency. This analysis assumes that the firm is permitted under MiCA Article 59 to provide custody and administration within Article 3(1)(17).

Illia Prokopiev

Stablecoin regulation in the US, Hong Kong and Singapore

Stablecoin and digital-token regulation now combines market-entry authorization with continuous financial-crime controls in daily operations. The question is whether the United States’ proposed payment-stablecoin customer identification program, Hong Kong’s narrow first licensing round, and Singapore’s digital payment token (DPT) directory support a bank-like compliance characterization. They do, with material limits. The more accurate proposition is that compliance is moving beyond approval into continuous financial-institution-grade operations.