LightGBM predict_proba in thousandths place

Can someone explain to me how my lightgbm classification model's predict_proba() is in thousandths place for the positive class:

prob_test = model.predict_proba(X_test)

print(prob_test[:,1])
    
array([0.00219813, 0.00170795, 0.00125507, ..., 0.00248431, 0.00150855,
       0.00185903])

Is this common/how is this calculated?

Should there be concern on performance testing(AUC)?

FYI: data is highly imbalanced train = 0.0017 ratio