EU AI Act Article 50: Complete Compliance Guide for Transparency Obligations

Article 50 transparency obligations apply from 2 August 2026. However, the Omnibus amendment grants providers of generative AI systems already on the EU market before that date a transitional period until 2 December 2026 to comply with the synthetic content marking requirement under Article 50(2). Systems placed on the market from 2 August 2026 onwards must comply immediately.

Both providers (developers) and deployers (business users) of AI systems have obligations under Article 50, but different ones. Providers must design systems to inform users of AI interaction and mark synthetic content in machine-readable format. Deployers must inform individuals about emotion recognition, biometric categorisation, and deepfake content. The obligations apply regardless of whether the provider is established in the EU, as long as the AI system is used in the EU.

Any AI system intended to interact directly with natural persons. This includes chatbots, virtual assistants, AI-powered customer service tools, AI agents, and conversational AI. The obligation does not apply if the AI nature is obvious from the circumstances, for example a clearly labeled chatbot widget on a website. The European Commission Guidelines clarify that the test is whether a reasonably well-informed, observant, and circumspect person would recognise they are interacting with AI.

Providers of AI systems that generate synthetic audio, image, video, or text must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. This applies to generative AI tools including large language models, image generators, voice cloning tools, and video synthesis systems. The Code of Practice recommends a multi-layer approach combining C2PA metadata and imperceptible watermarking such as Google SynthID.

Yes, but with nuance. Providers must mark AI-generated text in machine-readable format (Article 50(2)). Deployers who publish AI-generated text on matters of public interest must disclose that the text was AI-generated (Article 50(4)). However, this deployer obligation does not apply if the content has undergone human review or editorial control and a person holds editorial responsibility for the publication.

Deployers of AI systems that generate or manipulate image, audio, or video content constituting a deepfake must disclose that the content has been artificially generated or manipulated. Exceptions exist for content that is part of an evidently artistic, creative, satirical, fictional, or analogous work, and for authorised law enforcement activities. The disclosure must not hamper the display or enjoyment of the work.

Deployers of emotion recognition systems or biometric categorisation systems must inform the natural persons exposed to them of the system's operation. They must also process personal data in accordance with GDPR (Regulation 2016/679). Note that emotion recognition AI is entirely prohibited in workplaces and educational institutions under Article 5(1)(f), except for medical or safety purposes.

Non-compliance with Article 50 transparency obligations can result in administrative fines of up to €15,000,000 or 3% of total worldwide annual turnover, whichever is higher. For SMEs and startups, the fine is capped at the lower of the two amounts (not the higher), providing relative protection for smaller operators. EU institutions face fines up to €750,000.

The Digital Omnibus agreement (signed 7 July 2026) deferred Annex III high-risk AI obligations by 16 months to 2 December 2027, and Annex I obligations to 2 August 2028. However, Article 50 transparency obligations remain on their original schedule of 2 August 2026. The only relief is a transitional period until 2 December 2026 for providers of generative AI systems already on the market before 2 August 2026 to comply with the content marking requirement.

On 10 June 2026, the European Commission published the final Code of Practice on Transparency of AI-Generated Content. It provides technical guidance for complying with Article 50(2) marking requirements. The Code mandates a multi-layer approach: at minimum, a combination of C2PA cryptographically signed metadata and imperceptible watermarking deployed simultaneously. Adherence creates a presumption of conformity but is voluntary.

Generally yes. While the AI Act provides certain exemptions for free and open-source AI models, the transparency obligations under Article 50 apply broadly. If an open-source model generates synthetic content, its provider must ensure outputs can be marked and detected as AI-generated. The key factor is whether the model is placed on the EU market or its outputs are used in the EU.

The European Commission Guidelines recommend clear and distinguishable disclosure at the latest at the time of first interaction. Practical implementations include a notice before conversation begins (e.g., 'You are chatting with an AI assistant'), a persistent label visible during interaction, and accessible formatting that meets WCAG standards. The disclosure must be proactive, not buried in terms of service.

While Article 50 does not prescribe specific record-keeping, the Commission Guidelines and Code of Practice recommend maintaining: an inventory of AI systems subject to transparency obligations, documentation of disclosure mechanisms implemented, evidence of machine-readable marking for synthetic content, training records for staff handling AI-generated content, and incident logs for transparency failures.

Article 50 operates alongside GDPR, not as a replacement. GDPR Article 22 gives individuals rights regarding automated decision-making. GDPR Articles 13-14 require disclosure of automated processing. Article 50 adds AI-specific transparency requirements on top. For emotion recognition and biometric categorisation, deployers must comply with both Article 50 disclosure and GDPR data processing requirements simultaneously.

Yes, through extraterritorial application. If a provider outside the EU places an AI system on the EU market, or if the output of an AI system is used in the EU, Article 50 obligations apply. This means US, UK, and Asian AI companies serving EU users must comply with chatbot disclosure, content marking, and other transparency requirements.

With enforcement beginning 2 August 2026, organisations should immediately: (1) Complete an AI system inventory identifying all systems subject to Article 50, (2) Implement chatbot and virtual assistant disclosure notices, (3) Begin deploying content marking for synthetic outputs, (4) Establish deepfake disclosure procedures, (5) Review and update privacy notices for emotion recognition systems, (6) Train staff on transparency obligations, and (7) Document all compliance measures for potential regulatory inspection.