The EU AI Act’s August 2, 2026 Deadline: Article 50 Transparency, Explained
Article 50 of the EU AI Act applies from 2 August 2026. If your product uses a chatbot or generates content with AI, here is what you actually have to do — in plain English.
The EU AI Act is being phased in over several years, and the phase that affects the most businesses arrives on 2 August 2026: Article 50, the transparency obligations. Unlike the Act's high-risk rules, which apply to a narrow set of systems, Article 50 is relevant to almost any business that uses generative AI to interact with people or produce content. If you have a chatbot or you publish AI-generated text, images, audio, or video, this applies to you.
What Article 50 requires
Article 50 splits obligations between providers (those who build or supply an AI system) and deployers (those who use one). The core requirements are straightforward to state:
- Chatbots must announce themselves. If an AI system is intended to interact directly with people, those people must be informed they are talking to an AI — unless it is already obvious from the context.
- AI-generated content must be marked. Providers of generative AI must mark outputs in a machine-readable format so they are detectable as artificially generated or manipulated.
- Deepfakes and manipulated media must be disclosed. Deployers who create or manipulate image, audio, or video content that resembles real people, objects, or events (a "deepfake") must disclose that it is artificially generated.
- AI-generated public-interest text must be disclosed. Deployers publishing AI-generated text to inform the public on matters of public interest must disclose that it was artificially generated, unless it went through human review with editorial responsibility.
- Emotion recognition and biometric categorization must be disclosed. People exposed to these systems must be told they are being used.
The disclosures must be clear, and provided at the latest at the point of first interaction or exposure.
Who is in scope
In practice, Article 50 reaches far more organizations than the Act's high-risk provisions. Any business running a customer-facing chatbot is in scope. Any business generating marketing copy, images, or synthetic voices with AI is in scope. The obligation is not limited to AI companies — it applies to the deployer using the tool, which could be any company on the web.
The timeline you need to know
The transparency obligations apply from 2 August 2026. There is one transitional easing worth noting: under the AI Omnibus arrangement, generative AI systems already on the market before that date have until 2 December 2026 to meet the machine-readable marking requirement. The disclosure obligations themselves — telling people they are talking to a bot, labeling deepfakes — still begin on 2 August. The European Commission has also been developing a Code of Practice on transparency; signatories can rely on adherence to it as a way to demonstrate compliance.
What to do before the deadline
- Audit where you use AI to interact or generate. List every chatbot, every AI content-generation feature, and any emotion-recognition or biometric use. Most companies underestimate how many they have.
- Add the disclosures. Make sure chatbots identify themselves, AI-generated public content is labeled, and any synthetic media is disclosed. These are usually small UI and copy changes.
- Handle machine-readable marking. For generative outputs, plan for the marking requirement (watermarking or provenance metadata), using the transitional deadline if your system predates August.
- Document it publicly. Put your AI transparency posture on your trust center — which systems are AI, how you disclose them, and how your data is handled. This both satisfies the spirit of Article 50 and answers the buyer questions that arrive alongside it.
Why this is also a sales opportunity
Article 50 forces a conversation about AI transparency that your customers are already having internally. Companies that get ahead of it — with clear disclosures and a public AI governance page — turn a compliance obligation into a trust signal. The deadline is a marketing hook as much as a legal one: "our chatbot tells you it's a bot, our AI content is labeled, and here's exactly how your data is handled" is a stronger position than scrambling in late July.
This article is a plain-English summary, not legal advice. The obligations depend on your specific AI systems and role; consult qualified counsel for your situation. The primary source is Article 50 of Regulation (EU) 2024/1689.