The European Union began enforcing its AI Act transparency rules on August 2, 2026, marking a significant shift from voluntary compliance to mandatory enforcement. According to recent reports, the European Commission's AI Office and national regulators gained enforcement powers for the first time, with the AI Office now able to demand documentation, evaluate models directly, order corrective measures, or pull models from the EU market. The enforcement powers apply to transparency breaches carrying fines of up to €15 million or 3% of worldwide turnover, while prohibited practices face penalties of up to €35 million or 7% of global annual turnover. This represents a fundamental change from the previous system where the AI Act operated largely on trust, with general-purpose AI model providers carrying documentation and copyright obligations since August 2025.
The Digital Omnibus amendment package signed July 8, 2026, significantly postponed high-risk AI system obligations that were previously scheduled for August 2. As reported by the European Commission, hiring, credit scoring, and law enforcement systems now have until December 2027 to comply with their obligations. AI embedded in regulated products, such as medical devices, has until August 2028 to meet compliance requirements. However, this extension does not suspend existing transparency requirements, with generative systems already on the market having until December 2, 2026, to add machine-readable watermarks to synthetic content. The Commission emphasizes that organizations should use this extension to assess their AI systems and implement the required governance, documentation, testing, and monitoring measures.
The European Commission groups the main transparency duties into three categories: deepfakes, AI-generated public-interest text, and disclosure that a user is interacting with an AI system such as a chatbot. According to the latest enforcement guidelines, AI systems that interact directly with people must reveal they are machines from the first interaction onward, covering chatbots, voice assistants, and agents. The rules apply to any provider or deployer whose system reaches users in the EU, regardless of where the company is based. AI-generated or manipulated images, audio, and video must be flagged as artificial, including photorealistic ads and promotional content with AI-generated voiceovers. Text published to inform the public needs a label unless a human editor has reviewed it and taken responsibility, while companies running emotion recognition or biometric categorization systems must inform every person exposed to them.
Non-compliance with Article 50 can lead to administrative fines of up to €15 million (about $17.08 million) or 3 per cent of global annual turnover of the AI provider or deployer, whichever is higher. According to the latest enforcement framework, there are limited exceptions where deepfakes used in clearly artistic, satirical, fictional or similarly creative works can use a lighter disclosure approach, so long as the notice does not spoil the experience. AI-generated text is exempt if it has undergone human review or editorial control and a legally responsible person or organisation accepts editorial responsibility. AI chatbots used for customer service must be explicitly disclosed on company websites from the implementation date. Personal, non-professional use remains excluded from the AI Act, with the situation changing only when content is used professionally or commercially by businesses, freelancers, or monetised influencers.
The AI Act applies to providers and deployers whose systems or services are placed on the EU market or used in the EU, with the impact likely to extend well beyond Europe. Meta and Google have signed the voluntary code, signalling their intent to align with the framework. The EU AI Act is the world's first comprehensive AI law, adopted in 2024, using a risk-based model where unacceptable-risk AI systems are banned and high-risk systems face strict compliance requirements. The Act's reach is extraterritorial, applying to any organization whose AI touches the EU market, regardless of where that organization is headquartered. The first enforcement actions will demonstrate how hard the AI Office intends to swing in implementing these new compliance requirements across the EU market.
