Educational Tool for EU AI Act Understanding
Key milestones and current obligations under the EU AI Act
GPAI Obligations (Aug 2, 2025): General Purpose AI model providers must comply with transparency, copyright, and safety requirements. Code of Practice provides compliance pathway.
High-Risk AI Systems (Aug 2, 2026): 281 days remaining for full compliance with risk management, conformity assessment, and documentation requirements.
Prohibited AI Practices (Feb 2, 2025): Bans on social scoring, real-time biometric identification, and exploitative AI systems are in effect.
Understanding the EU's four-tier risk classification system for AI regulation
Do this quick check to get a preliminary understanding if and how the AI Act may apply to your AI system or use case
Obligations active since August 2, 2025 • Code of Practice published July 2025
Comprehensive documentation including model architecture, training process, data sources, and computational resources.
Information for downstream providers and public summary of training data using EU standardized template.
Policies to respect intellectual property rights and comply with copyright law in training data usage.
Additional requirements for the most advanced general purpose AI models with systemic risk capabilities
The GPAI Code of Practice, published July 10, 2025, provides a voluntary but highly recommended pathway for demonstrating compliance. Signatories benefit from:
AI systems that significantly impact health, safety, or fundamental rights requiring comprehensive compliance
Comprehensive obligations coming into force August 2, 2026 for all high-risk AI systems
| Requirement | Description | Key Components | Article Reference |
|---|---|---|---|
| Risk Management System | Continuous iterative process throughout the AI system's lifecycle to identify, estimate, and evaluate risks. | Risk identification, assessment, mitigation measures, regular monitoring | Article 9 |
| Data Governance | Ensuring training, validation, and testing data sets are relevant, representative, and free from errors. | Data quality management, bias mitigation, gap assessment | Article 10 |
| Technical Documentation | Comprehensive documentation of the AI system's development, functionality, and performance. | System description, algorithms, training methodologies, risk measures | Article 11, Annex IV |
| Human Oversight | Ensuring humans can understand, monitor, and intervene in AI system operations. | Understanding capabilities, anomaly detection, intervention capabilities | Article 14 |
| Conformity Assessment | Procedures to demonstrate compliance with AI Act requirements before market placement. | Internal control, third-party assessment, CE marking, EU database registration | Articles 43-46 |
Essential terms and concepts from the EU AI Act with official references
Key dates and milestones for EU AI Act compliance with current status updates
The EU AI Act officially entered into force, beginning the phased implementation timeline for various AI system categories.
Bans on unacceptable risk AI systems came into effect, including social scoring, real-time biometric identification in public spaces, and exploitative AI systems.
The General Purpose AI Code of Practice was published, providing voluntary compliance guidance for GPAI model providers with presumption of conformity benefits.
Obligations for general purpose AI model providers became applicable, including transparency, copyright, and safety requirements for systemic risk models.
Full obligations for high-risk AI systems will apply, including risk management, conformity assessment, technical documentation, and human oversight requirements.
GPAI models placed on the market before August 2, 2025 must achieve full compliance with all applicable requirements.