Since August 2, 2026, a new phase of the European regulation on artificial intelligence (AI Act) has come into effect. Article 50 of this text now imposes transparency obligations on all generative AI systems accessible in the European Union, whether or not they are classified as “high-risk.” Chatbots, image generators, audio or video generators: nearly all consumer AI tools are affected.
What Article 50 actually changes
The European text distinguishes several obligations depending on the type of system. Conversational agents (chatbots, voice assistants) must now clearly inform the user that they are interacting with an artificial intelligence rather than a human being, except when this is already obvious from the context. Providers of content generators, for their part, must mark the outputs of their tools in a machine-readable format, in order to facilitate their later detection by platforms.
The case of deepfakes is subject to a specific rule: any image, audio, or video generated or modified by AI that resembles a real person, place, or event, to the point of being mistaken for authentic, must carry a visible or audible notice disclosing its artificial origin. Notably, the European Commission specifies that it is not necessary to prove an intent to deceive for the obligation to apply.
A grace period for technical labeling
Not all obligations take effect at the same pace. While displaying a notice readable by the user is required starting August 2, automated technical labeling (digital watermarking) benefits from a tolerance period until December 2026 for tools already on the market before that date, giving time for harmonized technical standards to be finalized.
Who enforces this, and what are the risks for companies?
Oversight of the system falls to the European AI Office, in coordination with the competent national authorities of each member state. Penalties for non-compliance can reach 15 million euros or 3% of the company's annual global revenue, whichever amount is higher. An important point for non-European companies: the rule applies to any company whose services reach users located in the European Union, regardless of where the company is established.
Why this milestone matters
Following the entry into force of rules on high-risk systems, this new deadline marks a turning point: this time it affects all consumer uses of generative AI, from virtual assistants to image-editing tools. For users, the stated goal is simple: to be able to distinguish, without specialized detection tools, content produced by a machine from authentic human content. For companies in the sector, this means a quick review of their interfaces and content production processes to come into compliance.
Want to follow regulatory developments affecting digital technology and the economy? Check out our analyses in the Tech / Net section and our economic breakdowns in Economy.
- Obligation to disclose the use of an AI chatbot to the user
- Mandatory labeling of deepfakes, with no need to prove intent to deceive
- Machine-readable technical labeling, with tolerance until December 2026 for tools already in service
- Penalties of up to 15 million euros or 3% of global revenue
To better understand the full European regulatory framework on digital technology, also check out our Consumer section.
Frequently Asked Questions
Does this apply to all artificial intelligence tools?
Not only systems classified as “high-risk”: Article 50 broadly applies to all generative AI systems that interact with the public or produce synthetic content, regardless of their risk level.
Does an American company have to comply with these rules?
Yes, as soon as its service reaches users located in the European Union, regardless of the country where the company is established.
What happens to a company that does not comply with these obligations?
It faces a fine of up to 15 million euros or 3% of its annual global revenue, whichever amount is higher, imposed by the competent authorities in coordination with the European AI Office.
Sources
- Travers Smith — Is it a bot? EU AI Act transparency rules take effect 2 August 2026
- IBTimes UK — Chatbots Must Now Confess They're AI and Deepfakes Need Labels Under New EU Rules
- EU Artificial Intelligence Act — Article 50: Transparency Obligations
Telemarketing Calls: What Really Changes on August 11, 2026
The phone ringing in the middle of the day for a “free energy audit” or an “exceptional offer” could soon be a thing of the past. Starting August 11, 2026, commercial telemarketing calls will be banned by default in France:…
Senatorial Elections: How the Renewal of the French Senate Works
On September 27, 2026, about 160,000 grand électeurs (special electors) will go to the polls to renew half of the French Senate. This election, often less covered by the media than the presidential or legislative races, nevertheless shapes the country's…
E-Invoicing, Sick Leave, Retirement: What's Changing for Businesses and Employees
September 1 marks a regulatory turning point for French businesses and for millions of employees. Widespread electronic invoicing, capped sick leave, a new way of calculating retirement benefits for parents, and a “polluter pays” fee on PFAS: several measures take…
Economic Simplification Law: What's Really Changing for Businesses
A new step in administrative simplification has just been taken in France. The economic simplification law, enacted on May 26, 2026 and already partly in force, continues to make headlines this summer as its first concrete measures take effect for…
This article was written with the help of artificial intelligence. Editorial policy
Table des Matières




