Article 50 of the EU AI Act has applied since August 2, 2026, and brands and agencies that publish AI-generated deepfakes or AI-written public-interest text to EU audiences must disclose it. The Digital Omnibus left those duties in place. It delayed only the machine-readable marking duty for AI tool makers already on the EU market, to December 2, 2026, according to Morgan Lewis.
Last updated October 9, 2026. This article is general information, not legal advice. Have EU counsel review your disclosure approach before you rely on it.
Who must do what under Article 50: providers or deployers?
Article 50 splits its duties by role. A provider is the organization that builds an AI system and puts it on the EU market under its own name. A deployer is an organization that uses someone else's AI system for professional purposes. Most brands and agencies are deployers of tools such as Midjourney, Adobe Firefly, Runway, ElevenLabs, ChatGPT or Claude.
| Role | Duty under Article 50 | Applies from |
|---|
| Provider | Tell people when they interact with an AI system, unless that is obvious | August 2, 2026 |
| Provider | Mark synthetic text, image, audio and video outputs in a machine-readable format | August 2, 2026 for new systems; December 2, 2026 for systems already on the EU market |
| Deployer | Disclose deepfakes at first exposure, in a clear and distinguishable way | August 2, 2026 |
| Deployer | Disclose AI-generated text published to inform the public on matters of public interest, unless a human reviewed it and holds editorial responsibility | August 2, 2026 |
| Deployer | Inform people exposed to emotion-recognition or biometric-categorisation systems | August 2, 2026 |
Morgan Lewis states that the December 2 transition period does not postpone deployer duties. The European Commission published guidelines on the Article 50 transparency obligations on July 20, 2026, and a separate Code of Practice on Transparency of AI-Generated Content. The legal text is on the Commission's AI Act Service Desk.
Does the EU AI Act apply to US brands and agencies?
The EU AI Act applies to US companies when their AI system's output is used in the EU, whether or not the company has a European office. Article 2 of the Act sets this reach for providers and deployers located outside the EU. A US brand that publishes an AI-generated deepfake to EU audiences on Meta, Google or TikTok is in scope, as is an agency that runs that campaign.
Map your EU exposure first. List every campaign, client account and content channel that reaches people in the EU, and the AI tools that produced or edited each asset.
When does AI-generated text need a disclosure?
AI-generated text needs a disclosure when it is published to inform the public on matters of public interest and no human has reviewed it and taken editorial responsibility. Legal summaries describe this as an editorial-control exemption: AI-assisted text that a named person reviewed and approved before publication does not need the label.
Treat that exemption as something you must prove. Record who reviewed each piece, when, and what authority that person had to change or reject it. Marketing copy that does not inform the public on a matter of public interest may fall outside this duty, so classify each text asset before you decide.
What does the December 2 deadline change for brands and agencies?
The December 2, 2026 deadline changes the duty of tool providers, not the duties of brands and agencies. It gives providers of generative systems already on the EU market until that date to add machine-readable marking to their outputs. Systems placed on the market on or after August 2 must mark outputs from the start.
For a deployer, this creates a vendor question. Ask every generative-tool vendor whether its outputs carry machine-readable marking, and whether it relies on the December 2 transition. Legal500 notes that deployers must not remove or alter marks that providers embed, so check that your export and editing steps leave them intact.
What should brands and agencies do before December 2?
| When | Action | Output |
|---|
| This week | Inventory every EU-facing asset and the AI tool behind it, then pause any deepfake without a disclosure | Asset and tool register |
| By end of October | Classify each asset as deepfake, public-interest text, or other, and set one disclosure format that stays visible on mobile | Disclosure standard, tested on TikTok, Instagram, YouTube, email and LinkedIn |
| By end of October | Name a human editor and record the review process for AI-assisted text that informs the public | Editorial control log |
| By mid-November | Send each vendor the marking and transition questions, and have counsel review contracts for compliance warranties and indemnities | Vendor status sheet |
| Before December 2 | Re-check vendor status and confirm your editing workflow keeps embedded marks | Signed-off compliance file |
Build disclosure into the creative brief and final quality check. A label added after an asset has gone live leaves a period of non-compliant exposure.
What are the penalties for breaking Article 50?
Breaches of Article 50 can be fined up to 15 million euros or 3% of worldwide annual turnover, whichever is higher, under the AI Act's penalty rules. National market surveillance authorities in each EU member state enforce the transparency duties. Smaller companies face the lower of the two figures.
The decision for a US brand or agency is whether every EU-facing AI asset can be traced to a tool, a reviewer and a disclosure. If it cannot, pause that asset for EU audiences until it can.
