Raw content stays out of storage
Only a sanitized preview, SHA-256 hash, risk score, and explanation are persisted.
SentinelAI gives every submission a 0–100 risk score, explains the strongest signals, and stores only privacy-safe metadata for your dashboard.
This MVP uses a local heuristic/NLP engine as the safe baseline before adding trained Scikit-learn models and public cybersecurity datasets.
Only a sanitized preview, SHA-256 hash, risk score, and explanation are persisted.
Every result lists weighted indicators so users can understand what triggered the score.
The scanner interface is ready to blend in trained models, evaluation metrics, and safe updates.