LexOculus

The tool connects to a GitHub repository and parses all dependency files, mapping each listed library against a curated database of more than two hundred AI‑related packages. It then performs a language‑model‑powered analysis of the codebase to identify the system’s capabilities—such as computer‑vision, natural‑language processing, biometric or generative AI functions—and assigns confidence scores to each detection.

Based on the identified capabilities, the engine matches the system to the relevant EU AI Act Annex III articles and determines a risk tier (unacceptable, high‑risk, limited‑risk, or minimal‑risk), producing a numeric risk score. A dynamic questionnaire further refines the assessment by incorporating deployment context, human‑oversight measures, and other safeguards.

Finally, the service generates a signed compliance report that documents the scan, capability detection, risk classification, and context verification, enabling teams that ship AI to demonstrate EU AI Act compliance on a pull‑request basis.