EU AI Act complianceAI safety regulationsData SciencePythonMLflowAutomated AI governanceModel validation
Building EU AI Act Compliance Pipelines with Python and MLflow
7 min read · 1,358 wordsBy Orandi Felix
The EU AI Act lists **23 mandatory technical measures** in Annex IV for high-risk AI systems. Below is only the subset I automated; the full translation matrix is in the repo’s `compliance_map.csv`.
I added a lightweight `Click` CLI front-end so engineers can trigger compliance checks from Jenkins/CD. The CLI accepts a `--risk-class` flag (`high`, `limited`, `minimal`) and tailors the validator list accordingly.
The PDF is minted with `reportlab` and stamped with **EU time servers** (`ntplib`) to prove no backdating. Auditors can diff SHA-256 hashes against the tracking server to verify tampering.