Reg. (EU) 2024/1689
What you need to know
The world's first comprehensive AI regulation. In force since 1 August 2024 and applied in phases: prohibitions from February 2025, GPAI from August 2025 and transparency (Art. 50) from August 2026. The Digital Omnibus postpones high-risk obligations to December 2027 (Annex III) and August 2028 (Annex I). Extraterritorial scope and penalties of up to EUR 35 million or 7% of global turnover.
What is the EU AI Act?
Regulation (EU) 2024/1689, adopted on 13 June 2024 and published in the Official Journal on 12 July 2024, is the world's first comprehensive legislation on artificial intelligence. It sets risk-based obligations for providers, importers, distributors and deployers of AI systems operating in the European market. Its scope is extraterritorial: if your AI system has effects on people in the EU, the Regulation applies regardless of where your company is incorporated.
The four phases of the EU AI Act
Phased implementation. Phases 1, 2 and 3 already apply (the third since 2 August 2026). The Digital Omnibus (Regulation (EU) 2026/1744) postpones high-risk obligations: Annex III to 2 December 2027 and Annex I to 2 August 2028.
The four risk levels
The AI Act classifies AI systems according to the risk they pose. The classification determines which obligations apply and from when.
Obligations for high-risk systems
Annex III systems must meet 10 main obligations by 2 December 2027, following the postponement introduced by the Digital Omnibus (Regulation (EU) 2026/1744). For AI embedded in regulated products under Annex I, the deadline is 2 August 2028. They apply to providers and deployers operating in the EU.
Mandatory transparency (Art. 50)
In force since 2 August 2026. Applies to conversational systems, content generators and emotion recognition systems. Generative AI systems placed on the market before that date have until 2 December 2026 to apply machine-readable marking (Art. 50(2)).
Conversational systems
Chatbots and voice agents must identify themselves as AI at the start of every interaction. Users must know they are not talking to a human being.
HT Kiire and HT Meel Voice, in development, are designed to complySynthetic content
AI-generated text, images, audio and video must be labelled as artificially generated content. Mandatory for generators and distributors.
Applies to any GPAI system in productionEmotion recognition
Deployers of emotion recognition systems must inform the people exposed to them. In workplaces and educational settings they are prohibited, except for medical or safety reasons (Art. 5).
Structural human-in-the-loop at HumanosTechAdministrative penalties
The AI Act sets three levels of penalties. Authorities take company size into account, with adapted caps for SMEs, startups and small mid-caps, but fines can still threaten the viability of a medium-sized company.
Supervisory authorities by country
Each Member State designates its own national market surveillance authority. The EU AI Office supervises GPAI models at European level. Select your country to see the competent authority, its designation status and the relevant official resources.
Five steps to prepare before December 2027
Whatever the size of your company, these five steps reduce regulatory exposure and provide a foundation for any compliance programme.
AI systems inventory
Identify every AI system in use, including those contracted from third parties and those used by teams without IT involvement.
Classification by risk level
For each system identified, determine whether it falls into the prohibited, high-risk, limited or minimal category according to the Regulation's criteria.
Conformity assessment
For high-risk systems, start the technical documentation, training data analysis and risk management in line with Art. 9.
Governance and training
Appoint internal owners for AI compliance. Adopt AI literacy measures (Art. 4) for those who operate or oversee AI systems: following the Digital Omnibus, the Regulation requires taking measures to support it.
Legal advice
Involve legal advisers specialised in European AI regulation to validate your strategy, especially for high-risk systems.
Structural compliance, not a configuration option
HumanosTech operates under Estonian legislation and European law (Humanos Tech OÜ, reg. 17179670). Our AI products are designed with EU AI Act compliance built into the architecture, not added afterwards.
HT Kiire
AI chat agent for technical support. It answers with your documentation, citing sources, and helps diagnose incidents. No impactful action is executed without explicit human confirmation. Waiting list open.
HT Meel Voice
Conversational voice agent for phone support in European environments. It will identify itself as AI at the start of every call (Art. 50) and hand the call to a human with full context. AI is processed in the EU. The voice part is the last stage of the plan; waiting list open.
HT BiasGuard EU
System for auditing and continuously monitoring bias in AI systems deployed in the EU. Designed to help deployers meet the data governance (Art. 10) and human oversight (Art. 14) obligations, with auditable reports and decision traceability.
Official sources and resources
All the information in this guide is based on verified primary sources. For compliance decisions, always consult the official text of the Regulation and the European Commission's guidelines.
Regulation (EU) 2024/1689, EUR-Lex
Full text in all official EU languages.
eur-lex.europa.euEU AI Act, official portal
Implementation, governance and up-to-date guidelines from the Commission.
digital-strategy.ec.europa.euEU AI Office
European AI Office: supervision of GPAI models and coordination between Member States.
EU AI OfficeCompliance Checker
Conformity assessment tool from the Future of Life Institute.
artificialintelligenceact.euMarket Surveillance Authorities
Up-to-date list of the authorities designated in each Member State.
National authoritiesAI Regulation in the BOE
Official publication of the Regulation in the Spanish Official State Gazette (BOE).
boe.esDo you need technical guidance on the EU AI Act?
We can help you with your AI systems inventory, risk classification and technical governance. Legal advice requires a specialised legal professional, and we will always say so clearly.