From August 2, 2026, the European Union’s Artificial Intelligence Act requires certain AI systems and content to be disclosed to people in the situations covered by Article 50. The goal is straightforward: reduce deception and manipulation so you can make more informed choices in a digital world filled with synthetic media. (source)
1. Direct AI interactions now need a clear signal. When you interact directly with an AI system, its provider generally must inform you no later than your first interaction or exposure. A customer-service chatbot is a clear example. The notice may not be necessary when the system’s AI nature is already obvious from the circumstances and context.
2. Synthetic media needs machine-readable marking. Providers of systems that generate or manipulate synthetic audio, images, video, or text must ensure their outputs carry a machine-readable mark that allows them to be detected as AI-generated or AI-manipulated. Deepfake material in advertising or corporate social media can therefore require disclosure, subject to exceptions and special treatment for artistic, satirical, fictional, and editorial contexts.
3. Some public-interest text has a separate disclosure rule. Deployers that publish AI-generated or AI-manipulated text on matters of public interest must disclose that use. The Act does not create a blanket requirement to label every AI-powered recommendation, ranking, internal message, contract negotiation, or business process. The relevant question is what the system does, who encounters its output, and which specific duty applies.
4. The rules matter to US companies serving people in the EU. The Act can cover providers that place AI systems or general-purpose AI models on the EU market, as well as providers or deployers outside the EU when the system’s output is used in the EU. A US company may therefore need to build notices and content-marking tools into services used by EU audiences.
There is a limited technical transition. AI systems placed on the market before August 2, 2026, have until December 2, 2026 to comply with the machine-readable marking and detection obligation. Content produced before August 2 does not need retroactive labeling. For everyone else, the practical change is already beginning: notices and labels should become easier to spot when an AI system is directly involved or when covered synthetic content is published.
Noncompliance can be expensive. Companies that fail to meet applicable obligations can face fines of up to €15 million or 3 percent of worldwide annual turnover, whichever is higher. That gives large international businesses a strong reason to review how their systems communicate with people and identify synthetic content.
For readers, the useful habit is simple: look for the disclosure when you are speaking with a chatbot or viewing covered synthetic media, and treat the label as information about how the content was made. For companies, the next step is to map EU-facing systems to the specific Article 50 duty that applies. Clearer signals should make it easier to judge what you are seeing, hearing, and being told online.
