On 20 July 2026, the European Commission published a set of guidelines to help companies meet the transparency rules in the EU AI Act. Those rules apply on 2 August 2026, a few weeks after the guidelines came out. If your first instinct is that a Brussels regulation is somebody else's problem, an ocean away, it is worth a second look. The rule was written to follow the output, not the office.
The AI Act can reach your company even if you have never set foot in Europe, and that is the core point for a US business owner. Article 2 of the Act applies to providers who place an AI system on the EU market, and to providers and deployers in a third country "where the output produced by the AI system is used in the Union." The scope line is blunt: it applies "irrespective of whether those providers are established or located within the Union or in a third country."
Translated for a P&L owner: if a customer in Germany chats with your AI support agent, or a client in France reads a report your AI wrote, the disclosure obligations can attach to you in Indiana.
The rule was written to follow the output, not the office.
What the transparency rules actually require
The guidelines interpret Article 50 of the AI Act, which covers AI systems that carry transparency risks. The obligations are specific and, for the most part, concrete enough to explain to your team in a single meeting.
If people talk to your AI, tell them
Article 50(1) says that providers of AI systems built to interact directly with people must make sure users know they are dealing with an AI. There is one carve-out: disclosure is not required when it is already "obvious from the point of view of a natural person who is reasonably well-informed." A support chatbot on your website is the clearest example. If a person could reasonably think they are messaging a human, the rule expects you to say otherwise.
If your AI makes content, mark it
Article 50(2) reaches providers of generative AI that produce synthetic audio, image, video, or text. Those outputs have to be "marked in a machine-readable format and detectable as artificially generated or manipulated." The Act asks for solutions that are "effective, interoperable, robust and reliable as far as this is technically feasible." That last clause is the Regulation's own acknowledgment that the marking technology is still maturing. It sets a direction of travel rather than a demand for perfection today.
If you publish or deploy, disclose
The deployer obligations are where most business owners will feel this first:
- Deepfakes. Article 50(4) requires deployers who generate or manipulate image, audio, or video deepfakes to "disclose that the content has been artificially generated or manipulated." A lighter disclosure standard applies to artistic, creative, satirical, or fictional work.
- Public-interest text. Deployers who publish AI-generated or AI-manipulated text on topics of public interest have to disclose it, unless that content went through human review with clear editorial responsibility.
- Emotion and biometric systems. Article 50(3) requires deployers of emotion-recognition or biometric-categorization systems to inform the people exposed to them.
The number that gets attention
Non-compliance with Article 50 sits in the middle tier of the AI Act's penalties. Under Article 99, fines can reach 15,000,000 EUR, or for a company, up to 3% of total worldwide annual turnover, whichever is higher.
of total worldwide annual turnover, or 15,000,000 EUR, whichever is higher. That is the maximum fine for a violation of the Article 50 transparency rules.
Source: EU AI Act, Article 99, 2024For context, the top tier reaches 35,000,000 EUR or 7% of turnover, reserved for the prohibited practices in Article 5. A lower tier of 7,500,000 EUR or 1% of turnover covers supplying incorrect or misleading information. Transparency violations fall in between, which tells you how seriously the regulators treat disclosure.
Most owners are not going to become the first enforcement case. The point is the direction. This is the first hard example many businesses will meet of AI disclosure moving from etiquette to law, with a real number behind it.
The voluntary shortcut, and its deadline
Alongside the rules, the European AI Office released a Code of Practice on Transparency of AI-Generated Content on 10 June 2026. It is voluntary. Signing it is a way to help demonstrate compliance with the marking and labelling obligations under Article 50(2), (4), and (5). In plain terms, a signatory has a practical tool for showing it meets those points.
The signing form deadline for the initial signatory list was 27 July 2026 at 18:00 CEST. That specific date is worth noting because it shows how fast this moved: the Code came out in June, the guidelines in mid-July, the obligations apply in early August. This is not a distant timeline.
Why this belongs on your radar even if you never sell into Europe
Two reasons, both practical.
First, the reach clause means the trigger is your output crossing into the EU rather than the location of your headquarters. A single European customer, contractor, or reader can pull you into scope. It is worth knowing where your AI-generated content actually travels.
Second, disclosure norms tend to move across markets. The habit of labeling AI work, telling people when they are talking to a machine, marking a generated image, disclosing an AI-written report on a public topic, is arriving as a broader expectation. The businesses that build the habit now will not have to scramble later.
This is where a company's AI capability shows up. In the 7 Levels of AI Proficiency, the early levels are about using the tools. The higher levels are about judgment: knowing where AI output goes, what it triggers, and what a responsible business does about it before a regulator or a customer asks. Disclosure discipline reads as a proficiency skill more than a legal chore. A team that already labels its AI work is operating at a higher level than one that treats disclosure as an afterthought.
Related reading: Level 3: The Lieutenant (Critical Thinker).
A next step you can take this week
Do a quick pass on where your business already uses AI in front of other people: customer-facing chat, generated images or video in marketing, AI-drafted content you publish, any tool that reads emotion or sorts people. For each one, ask a simple question: would a reasonable person know AI was involved? Where the answer is no, add a plain label or a line of disclosure. It costs almost nothing, it builds the habit before the norm hardens in your own market, and it puts your team on the right side of a rule that is already law somewhere your customers live.
Sources
- Guidelines on transparency obligations for providers and deployers of AI systems (European Commission)
- Regulatory framework for AI (European Commission)
- AI Act Article 50: Transparency Obligations (Regulation (EU) 2024/1689 text)
- AI Act Article 2: Scope (Regulation (EU) 2024/1689 text)
- AI Act Article 99: Penalties (Regulation (EU) 2024/1689 text)
- How to sign the Code of Practice on Transparency of AI-Generated Content (European Commission)
Frequently Asked Questions
Does this apply to my US company if I have no European operations?
It can. Article 2 extends the AI Act to providers and deployers in a third country where the output of the AI system is used in the EU. Your location does not decide it; where your output is used can.
When do the transparency rules start?
The Article 50 transparency obligations apply on 2 August 2026. The AI Act itself entered into force on 1 August 2024 and becomes fully applicable on 2 August 2026, with some exceptions.
Is a support chatbot covered?
If it is built to interact directly with people, Article 50(1) expects users to be told they are dealing with an AI, unless that is already obvious to a reasonably well-informed person.
Do I have to sign the Code of Practice?
No. It is voluntary. Signing is a way to help demonstrate compliance with the marking obligations in Article 50(2), (4), and (5), which can make showing compliance simpler. The initial signing deadline was 27 July 2026.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.