EU AI Act Article 50: Do Its AI Labelling Rules Reach a Canadian or US Business?

The short answer

Yes, EU AI Act Article 50 can apply to a Canadian or US business, because the Act covers providers and deployers outside the EU when the output of their AI system is used in the EU. If people in Europe chat with your website bot or see your AI-generated video or images, the chatbot-disclosure and deepfake-labelling duties that have applied since 2 August 2026 are worth taking seriously. This is general information, not legal advice.

What EU AI Act Article 50 actually requires

Article 50 is the transparency chapter of the EU's Artificial Intelligence Act (Regulation (EU) 2024/1689). It splits duties between providers (whoever builds an AI system or puts their name on it) and deployers (businesses that use one professionally). In plain English, the text of Article 50 asks for four things:

• Chatbots and voice agents (50(1)): providers must design systems that talk to people so those people are told they are dealing with AI, unless that is obvious.
• Machine-readable marking (50(2)): providers of systems that generate synthetic audio, images, video or text must mark the output so it can be detected as AI-generated.
• Emotion recognition and biometric categorisation (50(3)): deployers must tell the people exposed to them.
• Deepfakes and public-interest text (50(4)): deployers must disclose that a deepfake is artificially generated or manipulated, and must disclose AI-generated text published to inform the public on matters of public interest, unless a human with editorial responsibility has reviewed it.

Disclosures must be clear, distinguishable and given at the latest at the first interaction or exposure. Creative, satirical or artistic work gets a lighter touch: you still disclose, but in a way that does not ruin the work.

Why a business outside the EU can be caught

The reach comes from Article 2 of the Act. Point (c) of Article 2(1) brings in providers and deployers established outside the Union "where the output produced by the AI system is used in the Union". The only general carve-out for deployers is purely personal, non-professional use.

So the useful question is not "are we an EU company?" It is "does anything our AI produces land in front of people in the EU?" A Toronto e-commerce brand that ships to France, a US software firm with German customers, or a Canadian agency running ads in Ireland can all answer yes.

A few realistic examples:

What you do	Your likely role	Article 50 duty in play
Run a website AI chatbot that EU visitors use	Deployer (the vendor is usually the provider)	Make sure users are told it is AI; check your vendor builds this in
Build your own AI assistant and offer it to EU customers	Provider	Disclosure by design (50(1)); marking if it generates content (50(2))
Publish an ad with a realistic AI-generated spokesperson	Deployer	Label it as a deepfake (50(4))
Post AI-written news-style articles on public issues, unreviewed	Deployer	Disclose AI generation unless a responsible human edits it (50(4))
Use AI to tidy grammar in a blog post	Deployer	Generally outside the marking rules for standard editing assistance

The timeline, including the Digital Omnibus changes

There was a lot of talk in early 2026 about the EU delaying the AI Act. Some of that happened, but not for transparency. The Digital Omnibus on AI was adopted as Regulation (EU) 2026/1744, dated 8 July 2026 and published in the Official Journal on 24 July 2026. It pushed the high-risk rules back to 2 December 2027 (Annex III systems) and 2 August 2028 (AI inside regulated products).

For Article 50, the Omnibus added one narrow transition: providers of generative systems already on the market before 2 August 2026 have until 2 December 2026 to meet the machine-readable marking duty in Article 50(2). Everything else in Article 50, including chatbot disclosure and deepfake labels, has applied since 2 August 2026. A Paul, Weiss client memo from August 2026 reads it the same way.

Codes, guidelines and the label icons

Two documents now fill in the detail. On 10 June 2026 the European Commission published a final Code of Practice on marking and labelling AI-generated content. It is voluntary, but it is the most straightforward way to show you meet the rules. The Commission then issued guidelines on Article 50 on 20 July 2026, according to the Paul, Weiss memo.

Three points from that guidance matter most for smaller firms:

• "Obvious" is read narrowly. A disclosure tucked into terms and conditions, or a vague "this site uses LLMs", is not enough. The bar is higher where children, older people or people with disabilities may be using the service.
• Deepfakes do not require intent to deceive. If the person or scene is plausibly real, it counts, photorealistic or not.
• The human-review exemption for text is demanding. A quick skim or spell-check is not editorial control. The reviewer needs relevant competence, and you should be able to show the workflow.

What it costs to ignore

Under Article 99, breaching Article 50 can bring fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and start-ups the cap flips to whichever is lower. Enforcement against a small business in another continent is not the most likely scenario. The bigger risk is commercial: EU partners, platforms and marketplaces are starting to ask suppliers to confirm they comply.

A practical plan for a Canadian or US business

1. Map where AI output meets EU users. Website chat, phone agents, email assistants, ad creative, product images, social video.
2. Fix the chatbot opening line. Have your bot or AI receptionist say it is an AI assistant in its first message or greeting, and keep it visible. That one change also lines up with several US state laws we cover in our guide to AI chatbot disclosure laws.
3. Label realistic synthetic media. If an ad shows an AI-generated person, voice or scene that could pass as real, add a visible label. The Commission has published EU icons you can use.
4. Ask your vendors about marking. If a tool generates images, video or audio for you, the provider carries the Article 50(2) marking duty. Get it in writing that they will meet the 2 December 2026 deadline.
5. Keep a human editor on public-interest text. If you publish commentary on public issues drafted with AI, document who reviewed it.

If your customers are mostly in Canada, the transparency habits above also line up with where Canadian law is heading. See our explainer on Bill C-36 and AI.

Key takeaways

• EU AI Act Article 50 reaches non-EU businesses when their AI output is used in the EU.
• Chatbot disclosure and deepfake labels have applied since 2 August 2026; the Digital Omnibus did not delay them.
• Only machine-readable marking for generative tools already on the market got extra time, to 2 December 2026.
• Disclose at the first interaction, in plain view, not in your terms.

Frequently asked questions

Does the EU AI Act apply to US companies?

It can. Article 2(1)(c) covers providers and deployers outside the EU where the AI system's output is used in the EU. A US company whose chatbot serves EU customers, or whose AI-generated ads run in the EU, should assume Article 50 is relevant.

Did the Digital Omnibus delay Article 50?

No. Regulation (EU) 2026/1744 delayed the high-risk rules, but Article 50 transparency duties have applied since 2 August 2026. The only extra time is for machine-readable marking by providers of generative systems placed on the market before that date, who have until 2 December 2026.

Do I have to label every AI-assisted image or blog post?

Not every one. Deployer labelling targets deepfakes and AI-generated text on matters of public interest that no responsible human has reviewed. Ordinary editing help and clearly fantastical images are treated differently. Provider-side machine-readable marking is the tool vendor's job.

Sources

1. Regulation (EU) 2026/1744 (Digital Omnibus on AI) — EUR-Lex, Official Journal, 24 July 2026
2. Article 50: Transparency Obligations for Providers and Deployers of Certain AI Systems — EU Artificial Intelligence Act (artificialintelligenceact.eu)
3. Article 2: Scope — EU Artificial Intelligence Act (artificialintelligenceact.eu)
4. Article 99: Penalties — EU Artificial Intelligence Act (artificialintelligenceact.eu)
5. Commission publishes Code of Practice on marking and labelling AI-generated content — European Commission, 10 June 2026
6. EU Finalises Transparency Rules for AI-Generated Content — Paul, Weiss, 4 August 2026

How Flowmatix can help

We put this into practice for businesses in Toronto and around the world through our AI Agents & AI Solutions, website design & development, content & video production services. Want to know what it would take for your business? Get a personalised quote in about 3 minutes.