About the AI Model Directory
The Complior AI Model Directory is an open catalog of AI models classified under the EU AI Act. Each model includes real metrics from HuggingFace — downloads, likes, parameters, license — alongside EU AI Act risk classification and deployer obligations.
For every model, we map the applicable EU AI Act articles, estimate compliance effort, and track deadlines across 108 mapped obligations.
Risk Classifications Explained
Prohibited — Tools banned under Article 5, including untargeted facial scraping and social scoring. Usage results in fines up to €35M.
High Risk — Tools used in critical domains: HR, credit, law enforcement, education, healthcare. Up to 19 deployer obligations including FRIA and conformity assessment.
GPAI — General-Purpose AI systems like Llama, Mistral, and GPT. 7 base obligations. Risk escalates based on deployment context.
Limited Risk — Tools with transparency obligations only (Art. 50). Includes chatbots, emotion recognition, and deepfake generators.
Minimal Risk — Most AI tools. No specific obligations, but voluntary codes of practice encouraged.