Your website chatbot answers pricing questions at 2 a.m. Your demand-gen team generates product visuals, a synthetic voiceover for the explainer video, and a first draft of every industry commentary post. None of that felt like a compliance question in July. Since August 2, 2026, parts of it are. The EU AI Act's transparency rules in Article 50 now apply and can be enforced by national authorities, and most of the headlines about the Act being delayed were about a different set of obligations entirely.
What is Article 50 of the EU AI Act?
Article 50 is the transparency chapter of the EU AI Act: it requires that people know when they are talking to an AI system and when content has been generated or manipulated by AI. It applies to any AI system that fits its criteria, regardless of risk classification, and it splits duties between two roles. A provider is whoever builds the AI system or places it on the market. A deployer is whoever uses it in a professional capacity, which is where most B2B marketing teams sit. According to Goodwin's August 2026 client alert, the transparency obligations became generally applicable and enforceable across the EU on August 2, 2026.
Wasn't the AI Act delayed?
Partly, and not for this. The Digital Omnibus on AI, which entered into force on July 27, 2026, postponed the high-risk system obligations: Gibson Dunn's analysis of the agreement puts the new Annex III deadline at December 2, 2027 and the Annex I deadline at August 2, 2028. Article 50 was left out of that deferral. The one targeted concession is narrow: generative AI systems already on the market before August 2, 2026 have until December 2, 2026 to meet the machine-readable marking requirement in Article 50(2). That grace period belongs to the providers of those systems, it covers marking only, and it does not extend any deployer duty. If your team read the word delayed and filed the topic under 2027, this is the correction.
The four obligations, translated into marketing work
Chatbots and AI assistants. Under Article 50(1), people must be clearly informed that they are interacting with an AI system, unless that is obvious from context. For a B2B team, that means the website chat widget, the AI sales assistant that books meetings, and any AI agent that replies to inbound email. The practical guide published by artificialintelligenceact.eu stresses that the information must arrive at the latest at the first interaction, which means before the conversation starts, not in a footer.
Machine-readable marking of synthetic content. Under Article 50(2), providers of generative AI systems must mark audio, image, video and text outputs in a machine-readable, detectable format. That duty sits with the model and tool vendors, not with you, but it has a marketing consequence: if your workflow strips metadata when you export, resize or compress assets, you may be quietly removing the marking your vendor was required to add.
Deepfakes. Under Article 50(4), deployers must disclose image, audio or video content that resembles real persons, places or events and could falsely appear authentic. A synthetic video of your CEO, an AI-cloned voice on a webinar promo, or a realistic generated image of a customer site all fall in scope. Evidently artistic, satirical or fictional work gets a lighter duty: disclosure in an appropriate manner that does not spoil the work.
AI-generated text on matters of public interest. Also under Article 50(4), deployers must disclose AI-generated or manipulated text published to inform the public on matters of public interest, unless it has gone through human review or editorial control and a person or company holds editorial responsibility. For most B2B content programs, that exception is the whole game: a named editor, a documented review step, and an owner of the final text.
What are the penalties, and who can enforce them?
Fines for breaching Article 50 reach up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher, according to Goodwin's summary. Enforcement sits with national market surveillance authorities in each member state. Being headquartered outside Europe does not take a company out of scope: the Act reaches providers and deployers in third countries where the output of the AI system is used in the Union. If your chatbot talks to buyers in Munich or your AI-generated video runs in a Paris campaign, assume the rules apply to you. None of this is legal advice; your counsel should confirm how each obligation maps to your specific systems.
The Code of Practice gives you a compliance path
The fastest way to operationalize Article 50 is the European Commission's Code of Practice on marking and labelling AI-generated content, published in final form on June 10, 2026. It is voluntary, but the Commission states that signatories can rely on its measures to demonstrate compliance with the rules on labelling AI-generated content, deepfakes and certain text publications. It includes guidance on where and how to show labels, including a publicly available EU icon for AI-generated content, and alternatives such as audio disclaimers where a visual label is impossible. Goodwin also reports that the Commission adopted its Guidelines on Transparency Obligations on July 20, 2026. Between the two documents, the question has shifted from what does the law mean to have we implemented it, which is a much better question for a marketing team to own.
Why this is a positioning signal, not only a risk
Transparency rules create a visible difference between vendors who disclose cleanly and vendors who get caught not disclosing. In B2B, where procurement and legal already sit inside the buying group, a clear AI disclosure on your chatbot and a documented editorial process for your content are credibility assets. They answer the vendor-risk questionnaire before it arrives. Companies that treat Article 50 as a marketing-operations standard rather than a legal footnote can say so in RFP responses and on their trust pages, while competitors are still discovering which of their tools generate content without a paper trail.
How Regulatory Radar and Policy Shield handle this
Robit Digital is not an agency and does not buy media; it runs the intelligence layer underneath marketing, with six AI agents monitoring a client's market around the clock and human strategists turning findings into written strategy. Agent 04 — Regulatory Radar tracks regulation, compliance deadlines, industry standards and grant windows in the client's sector, which is how an August 2 applicability date, a December 2 grace period and a July 20 guidance document land on a marketing calendar instead of in a post-mortem. Agent 06 — Policy Shield watches advertising-platform policy on Meta, Google and TikTok, runs pre-flight checks on campaigns and raises account-health alerts, because platform rules on synthetic and altered media evolve alongside the law, and a creative that satisfies one can still trip the other. Agent 01 — Competitor Intelligence shows how competitors are disclosing, or not, across their ads and channels.
Agent 02 — Human Strategist then turns those findings into the relevant parts of the 13 strategy documents, from messaging architecture to the positioning that makes transparency a selling point, reviewed at a bi-weekly strategic board and delivered in whatever language the team works in, at no extra cost. Regulatory Radar and Policy Shield are information and monitoring services, not legal or compliance advice; their job is to make sure the right change reaches the right person in time to act on it.
What to do this week
One: inventory every place AI touches a buyer — chat widgets, AI sales assistants, automated email replies, generated images, synthetic voice and video — and name an owner for each. Two: add a clear AI disclosure at the start of every chatbot conversation, not in the privacy policy. Three: write down your editorial review step for AI-assisted content, with a named editor, so the editorial-responsibility exception actually applies to you. Four: check whether your asset pipeline strips metadata on export, and fix it before December 2. Five: flag any synthetic depiction of real people, including your own executives, for explicit labelling, and read the Code of Practice section on the EU icon before your next campaign ships.
Article 50 is not a future problem and it was not postponed. It is a live rule that rewards teams who knew the date before it arrived. If you want regulation and platform policy in your category to reach you as a plan rather than a surprise, book a 20-minute discovery call and we will walk you through what Regulatory Radar is already tracking.
