Article 50 is in force: 5 things to do this week

Since 2 August 2026, the transparency obligations of Article 50 of Regulation (EU) 2024/1689 are law across the EU — for chatbots, AI-generated content, deep fakes and emotion-recognition systems. If your team spent last month hoping for another delay: the Digital Omnibus moved the high-risk rules, not Article 50. Here are the five things worth doing in the first week — each one concrete, each one backed by the official texts.

1. Confirm which paragraphs actually apply to you

Article 50 is five distinct duties, and most products trigger only one or two: direct-interaction disclosure (50(1)), machine-readable marking of synthetic content (50(2)), emotion-recognition and biometric-categorisation disclosures (50(3)), and visible deep-fake / public-interest-text labelling (50(4)). The European Commission's FAQ on Article 50 is the best official primer; our product-type breakdown sorts the duties by what you ship. The fastest way to a written answer: the free 3-minute checker maps your product to the exact paragraphs and gives you a dated gap report.

2. Ship the chatbot / voice-agent disclosure (Article 50(1))

If people interact directly with your AI — support chatbot, sales assistant, voice agent — they must be informed they are dealing with an AI system, unless that is obvious from the context. This is the cheapest duty to close: a clear line in the interface or greeting. Working examples in chatbot disclosure examples and voice-agent disclosure; a ready-made, 24-language disclosure badge is one script tag away.

3. Start machine-readable marking — do not wait for December

Providers of systems generating synthetic audio, image, video or text must mark outputs in a machine-readable, detectable format (Article 50(2)). Systems placed on the market on or after 2 August 2026 must mark from day one; systems already on the market before that date have a transitional window until 2 December 2026 (live countdown here). Four months is not much for a pipeline change: the implementation checklist compares C2PA Content Credentials, invisible watermarks and metadata per content type, and you can sign a test image with our C2PA API this week.

4. Label deep fakes and public-interest AI text visibly

Deployers publishing deep fakes must disclose that the content is artificially generated or manipulated; the same applies to AI-generated text published to inform the public on matters of public interest, unless it went through human editorial review (Article 50(4)). This is a visible label, distinct from the machine-readable mark — how to label AI content walks through placements that satisfy the clear-and-distinguishable standard of the Commission's transparency guidelines.

5. Document everything you just did

Enforcement sits with national market surveillance authorities (Commission overview of the enforcement framework), and fines for Article 50 breaches can reach EUR 15 million or 3% of worldwide turnover (Article 99(4)(g)) — but the realistic first question, from a regulator or an enterprise customer, is “show me what you did and when”. Keep a dated record of your assessment, the disclosures you shipped and the marking decisions you made: the evidence-pack guide covers what to keep. Consultancies doing this for clients can start from a white-label sample report (agency plan).

The 10-minute version

If you only do one thing today: run the free checker. It asks ~10 questions, cites the exact paragraphs that apply to you, and produces a dated gap report you can file as the first entry of your compliance record. Everything above links back to the official texts on EUR-Lex and the European Commission's Article 50 pages — read them, not the panic posts.