diff --git a/CHANGELOG.md b/CHANGELOG.md
index ee27515..13f9664 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,11 @@ Format: [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); versioning: [S
 
 ## [Unreleased]
 
+### Fixed
+- `features.feature_importance` (MDA): feature shuffling mutated a column view
+  in place, which raises `ValueError: array is read-only` under pandas
+  copy-on-write (default since pandas 3.0). Now shuffles a copy and assigns back.
+
 ## [1.0.9]
 
 ### Fixed
diff --git a/RiskLabAI/features/feature_importance/feature_importance_mda.py b/RiskLabAI/features/feature_importance/feature_importance_mda.py
index 20e63bf..0f839f0 100644
--- a/RiskLabAI/features/feature_importance/feature_importance_mda.py
+++ b/RiskLabAI/features/feature_importance/feature_importance_mda.py
@@ -97,7 +97,12 @@ def compute(self, x: pd.DataFrame, y: pd.Series, **kwargs: Any) -> pd.DataFrame:
             rng = np.random.default_rng(self.random_state + i) 
             for feature in x.columns:
                 x_test_shuffled = x_test.copy(deep=True)
-                rng.shuffle(x_test_shuffled[feature].values) # <-- USE SEEDED SHUFFLE
+                # Shuffle a copy and assign back: with pandas copy-on-write
+                # (default since pandas 3.0), column .values is a read-only
+                # view and in-place shuffling raises ValueError.
+                shuffled_values = x_test_shuffled[feature].to_numpy(copy=True)
+                rng.shuffle(shuffled_values)  # seeded shuffle
+                x_test_shuffled[feature] = shuffled_values
 
                 shuffled_proba = fitted_classifier.predict_proba(x_test_shuffled)