BRUSSELS, BELGIUM / RankWire.AI / – Starting August 2, 2026, new European Union transparency standards for artificial intelligence went into effect. According to Article 50 of the EU AI Act, companies now have clear obligations regarding the identification of certain AI-driven interactions and synthetic materials. The legislation encompasses chatbots, deepfakes, generated media, and specific public-interest texts. However, it does not mandate visible labels for all content produced via AI. Instead, each obligation is tailored to the system, the content produced, and how an organization presents it.
By mid-2026, companies utilizing interactive AI systems are required to disclose to users when they are engaging with artificial intelligence. This requirement is waived when a typical user can easily recognize the system’s artificial nature. Providers of generative AI are also mandated to include machine-readable data within synthetic texts, images, audio, and video. This information aims to facilitate automated detection of altered or generated content. The obligation applies when the technology can reliably and effectively support such marking.
Separate rules exist for organizations distributing synthetic media. They must clearly identify deepfake images, audio, or video that could be mistaken for real. Additionally, AI-generated texts about public-interest topics must be labeled in specific scenarios. This labeling requirement is not applicable if individuals review the material and exercise editorial control. Furthermore, a designated individual or organization must assume responsibility for publication before this exemption is granted.
Focus on Public Disclosures for Deepfakes and Civic Content
Furthermore, the rules extend to emotion recognition and biometric categorization systems that evaluate individuals. Operators must notify those affected when these technologies are employed, except in limited cases authorized by law. The EU AI Act provides more leniency for creative, fictional, artistic, and satirical works, allowing disclosures that do not disrupt their normal presentation or enjoyment. Nonetheless, operators must still provide proper notices if the work contains deepfake content.
In addition, the European Commission issued guidelines clarifying the scope of Article 50 and its exceptions. It also released a voluntary code designed for companies involved in developing or deploying generative AI. This code offers a framework for demonstrating compliance with labeling and marking standards. Those who opt out of the code must rely on alternative effective measures to ensure compliance. While optional icons can raise public awareness, they do not inherently guarantee legal adherence.
Authorities Hold Power to Enforce Heavy Penalties
Enforcement will primarily be managed by national market surveillance authorities across the EU. The AI Office will oversee a select group of systems related to general-purpose AI models and major platforms. Meanwhile, the European Data Protection Supervisor is tasked with monitoring relevant systems employed by EU institutions and agencies. Penalties for non-compliance can reach up to 15 million euros or 3% of a company’s global annual turnover. Regulatory bodies are authorized to impose these fines for violations.
For generative AI systems already on the market before August 2, a transitional period extends until December 2, 2026. During this time, providers must adhere to the machine-readable marking requirement. This extension solely concerns the technical aspect for detectable synthetic content, not the notices for chatbots, deepfake disclosures, or public-interest texts. Additionally, organizations are not mandated to label content created before the new transparency rules were introduced.