The European Union has moved a new part of its AI Act from policy into practice. As of August 2nd, fresh transparency obligations are in effect for companies that build, market or use AI systems.
The rules are aimed at a practical problem: people online can increasingly encounter chatbots, synthetic media and altered content without knowing what they are looking at or interacting with. The new requirements are designed to make those AI interactions and AI-generated materials easier to identify.
What Changed On August 2nd
The new EU AI Act transparency rules require companies to disclose when people are interacting with AI models. They also require disclosure when content has been generated or altered by AI.
That matters most in contexts where the system or content could be mistaken for something human-created or authentic. The rules cover both interactive AI systems and synthetic content, including audio, image, video and text.
The European Commission described the reason for the rules in its transparency guidelines, saying that generative and interactive AI systems are making it harder to tell AI interactions and AI-generated content apart from human-created and authentic content.
The Commission’s stated goal is straightforward: people should know when AI is involved so they can make informed decisions, judge how much trust to place in what they see, and avoid misinformation or deception.
Providers And Deployers Have Different Duties
The AI Act transparency rules separate responsibilities between providers and deployers.
Providers are companies that develop and market AI systems. Deployers are platforms and services that use those AI systems. Some companies can fall into both categories, with Meta and SpaceXAI named as examples in the source article.
For providers, the rules focus on the design of the AI system itself. They must build systems that clearly tell users when they are dealing with AI rather than a human, unless this is obvious. They also need to include machine-readable marks on synthetic audio, image, video and text so artificially generated or manipulated material can be detected.
For deployers, the emphasis is on labeling realistic deepfake content. Platforms and services using AI systems must label AI-generated or manipulated image, audio and video deepfakes when that content is designed to look real.
In plain terms, the EU is assigning responsibility at more than one point in the chain. The company that makes an AI system has to support transparency through the system’s design. The company that uses AI-generated or manipulated media in a service has to make sure users can see that AI was involved.
Labels Are Optional In Design, Not In Requirement
The European Commission has created a set of AI disclosure labels that tech platforms can adopt under the AI Act rules. It has also provided examples showing when those labels should be used.
Those EU-made icons are not the only possible way to comply. The source article notes that the examples are optional, and that the labels mirror similar disclosure labels already introduced by TikTok, Instagram and Facebook.
The point of providing EU-made icons is to reduce the need for every platform to design its own system from scratch. A shared set of examples can make compliance easier for companies and recognition easier for users.
But the Commission also made clear that the underlying labeling requirements are not optional. Platforms may be able to choose how they present disclosures, but they still need to meet the transparency obligations set by the AI Act.
What Companies Risk If They Ignore The Rules
The rules carry financial consequences. Companies that fail to comply with the new transparency obligations risk fines of up to €15 million (about $17.2 million), or 3 percent of their global annual turnover.
Compliance timing depends on when the AI system or service launched. For new AI systems, the transparency rules are enforceable immediately. For models and services that launched before August 2nd, there is a four-month grace period that runs until December 2nd.
That grace period gives existing systems time to adjust. It does not remove the obligation. After that deadline, older models and services are also expected to meet the same transparency rules.
Why The Rules Matter For Everyday Users
The EU’s approach is not framed as a ban on AI-generated content or chatbots. It is a disclosure system. The central idea is that users should not have to guess whether they are speaking to a human, seeing a real image, or encountering media that has been generated or altered by AI.
For chatbots, that means clearer notice when the user is interacting with AI. For synthetic media, it means labels and machine-readable signals that can help identify content created or changed by AI systems.
This is especially relevant for deepfakes designed to look real. If an image, audio clip or video appears authentic but was generated or manipulated by AI, the new rules require deployers to label it.
The result is a more formal transparency layer for AI in Europe. Companies still have choices about presentation, including whether to use the European Commission’s own disclosure labels. But the broader direction is now enforceable: AI systems and AI-made content need to be identified more clearly when users encounter them online.