Law explained

EU AI Act Article 50, explained in plain language (2026)

Article 50 of the EU AI Act is the “transparency” article: the set of rules that says people shouldn’t be deceived about when they’re dealing with AI or AI-made content. It has applied since 2 August 2026. Here is the whole article, duty by duty, in plain words.

50(1) — Tell people they’re talking to AI

If you run an AI system that interacts directly with people — a chatbot, a voice assistant — it must be designed so people are informed they’re dealing with AI, unless that’s already obvious to “a natural person who is reasonably well-informed, observant and circumspect.” In practice: a clear notice at the start of the conversation. Who owes it: the provider designs for it; if you deploy a chatbot on your site, you’ll want the notice visible regardless — you’re the face the customer sees. Full guide →

50(2) — Mark synthetic content machine-readably

Providers of AI that generates synthetic audio, images, video or text must ensure outputs are marked in a machine-readable way as artificially generated. Marking must be effective, interoperable, robust “as far as technically feasible.” Machine-readable marking applies since 2 August 2026; tools already on the market before that date have a transitional period until 2 December 2026 (2026 AI Omnibus). If you’re a small business using third-party AI tools, this duty belongs mostly to your tool’s maker — but you should choose tools that mark, and avoid stripping marks when editing. Full guide →

50(3) — Emotion recognition and biometric categorization

If you deploy these, you must inform the people exposed to them. Bigger picture: several such practices are outright prohibited (Art. 5) or high-risk (Annex III). If this is you, you need a lawyer, not a widget — we say this plainly.

50(4) — Label deepfakes and AI public-interest text

Deployers must visibly disclose (a) AI content that resembles real persons, places or events and could falsely appear authentic (“deepfakes” — even benign ones), and (b) AI-generated text published to inform the public on matters of public interest, unless a human editorial process with responsibility stands behind it. Deepfakes → · AI text →

50(5) — Timing and clarity

The information above must be provided clearly and distinguishably, at the latest at the first interaction or exposure, and meet accessibility requirements.

What Article 50 is not

It is not the “high-risk” regime (that’s Chapter III, phasing in through August 2028), and it’s not the GPAI-model rules for AI companies. For a typical small business, Article 50 is the part of the AI Act you actually touch.

Fines for violations: up to €15,000,000 or 3% of worldwide turnover, whichever is higher — with explicit proportionality for SMEs (Art. 99). The real numbers →

Sources are cited in place; official text: Regulation (EU) 2024/1689. This guide is a technical explanation, not legal advice.