EU begins enforcing landmark AI Act, sets global benchmark for responsible AI

From deepfake labelling and chatbot disclosures to oversight of advanced AI models, the legislation establishes a comprehensive compliance framework to address emerging risks while promoting responsible innovation across the digital ecosystem.

The European Union has officially entered a new phase of artificial intelligence regulation, with the first major provisions of its Artificial Intelligence (AI) Act coming into force on August 2. The move marks the beginning of enforcement of what is widely regarded as the world's first comprehensive legal framework governing AI, setting new compliance requirements for developers, technology providers and businesses operating across the bloc.

Originally proposed by the European Commission in 2021, the legislation underwent years of negotiations before receiving formal approval from the Council of the European Union in 2024. The framework establishes a common set of rules aimed at ensuring AI technologies are developed and deployed responsibly while safeguarding public safety, fundamental rights and consumer trust. Similar to the global influence of the General Data Protection Regulation (GDPR), the AI Act is expected to shape regulatory approaches well beyond Europe.

Transparency rules take effect for AI-generated content

One of the immediate requirements under the legislation is greater transparency around AI-generated content and interactive AI systems. Organisations deploying chatbots and conversational AI tools must clearly inform users when they are communicating with artificial intelligence rather than a human.

The rules also introduce mandatory disclosure requirements for AI-generated or AI-manipulated media, including images, audio recordings and videos. Such synthetic content, including deepfakes, must carry visible labels along with machine-readable identifiers to make automated detection easier across digital platforms. European regulators believe these measures will help curb misinformation, reduce online manipulation and improve public confidence in AI-enabled services.

Coinciding with the rollout, the European Commission also announced that more than 180 organisations have signed its voluntary Code of Practice supporting transparency for AI-generated content.

Risk-based oversight and tougher compliance measures

At the heart of the AI Act is a four-tier risk classification system that determines regulatory obligations based on the potential impact of an AI application.

AI systems considered to pose an unacceptable risk-including social scoring, certain forms of behavioural manipulation, predictive policing, emotion recognition in schools and workplaces, and most real-time biometric surveillance in public spaces-are prohibited, subject to limited exceptions.

Applications classified as high risk, including those used in healthcare, transport and law enforcement, will be required to meet stringent standards relating to testing, human oversight, risk management and data governance before they can be deployed within the EU market.

Lower-risk AI applications, such as chatbots and content-generation tools, remain permissible but must comply with transparency obligations, while minimal-risk applications, including spam filters and AI-enabled video games, are largely exempt from regulatory restrictions.

The legislation also places new responsibilities on developers of general-purpose AI models, particularly those capable of performing a wide range of advanced tasks. These providers must maintain detailed technical documentation, comply with copyright obligations and publish summaries of the data used to train their models. Models posing systemic risks—including threats linked to cybersecurity, critical infrastructure or public safety—will face enhanced regulatory scrutiny.

Enforcement responsibilities will be shared between the European AI Office, national regulatory authorities and the European Data Protection Supervisor, depending on the category of AI system involved. To strengthen implementation, the Commission has appointed Professor Alessandro Abate of the University of Oxford as Lead Scientific Adviser and established a Scientific Panel comprising 60 independent AI experts. The Commission has also introduced dedicated complaint and whistleblower mechanisms to support compliance monitoring.

Non-compliance with the AI Act could result in significant financial penalties linked to a company's global annual turnover, although proportionate limits have been предусмотрed for small businesses and start-ups. The remaining provisions of the legislation will be introduced in phases through 2028, including stricter rules for high-risk AI systems and AI integrated into regulated products, reinforcing the EU's long-term vision for trustworthy and accountable artificial intelligence.