The EU Now Requires Companies to Label AI Chatbots and Deepfakes
New rules that took effect on 2 August 2025 force any company operating in Europe to tell users when they are talking to an AI, or viewing AI-generated content. Fines of up to €15 million are on the table for those that ignore them.
Key points
- The EU's AI Act transparency rules came into force on 2 August 2025, requiring clear disclosure whenever users interact with an AI or see AI-generated content.
- Companies that develop AI systems must embed machine-readable markers in synthetic audio, images, video and text so the content can be detected as artificial.
- Platforms that deploy AI must label deepfakes, meaning AI-generated or AI-altered video, images and audio designed to look real.
- Non-compliance carries fines of up to €15 million, or 3 percent of a company's global annual revenue, whichever is larger.
- AI systems already live before 2 August have a grace period until 2 December 2025 to meet the new rules.
From 2 August 2025, any company offering AI products or services inside the European Union must tell users, clearly, when they are talking to a chatbot rather than a person. The same rules require that AI-generated images, videos and audio carry labels or hidden markers flagging them as artificial. The requirements come from the EU's AI Act, the bloc's sweeping law regulating artificial intelligence.
What exactly do the new rules require?
Two groups of companies face different obligations. Providers, the companies that build and sell AI systems, must design those systems to notify users whenever they are dealing with AI and not a human, unless the fact is already obvious. They must also embed machine-readable marks, invisible digital tags the way a watermark works on paper, inside any synthetic audio, image, video or text their systems produce.
Deployers, the platforms and services that plug existing AI into their products, must label any deepfake content, AI-generated or AI-altered video, images and audio made to look genuine, that appears on their services.
Some companies fall into both camps. Meta, which builds its own AI and also runs Facebook and Instagram, and xAI, Elon Musk's AI company behind the Grok chatbot, are each classified as both a provider and a deployer under the rules.
Why does the EU say this matters?
Ordinary people are losing the ability to tell what is real. The European Commission put it plainly in its guidelines: "The rapid development of generative and interactive AI systems is making it increasingly difficult to distinguish AI interactions and AI-generated content from human-created and authentic content."
The Commission has created a set of optional disclosure icons that platforms can adopt. The icons mirror labels already used by TikTok, Instagram and Facebook. The EU built its own set to spare companies from inventing competing symbols. The labels themselves are optional; the underlying legal obligation to disclose is not.
What should readers watch for?
If you are in Europe, look for small labels on videos and images that read something like "AI-generated" or carry a disclosure icon. When chatting with a customer-service bot or a virtual assistant on any major platform, you should now see a clear statement that the other side is not human.
If a video or voice clip seems designed to put words in a real person's mouth and carries no label, that is a warning sign worth acting on: reverse-image search the still frames, check the source, and do not share it before you do.
Companies that launched AI products before 2 August 2025 have until 2 December 2025 to comply, as first reported by The Verge AI. After that date, the fines become real.
