diff --git a/python-package/lightgbm/basic.py b/python-package/lightgbm/basic.py
index 3342e5205393..5c7d57bdd61b 100644
--- a/python-package/lightgbm/basic.py
+++ b/python-package/lightgbm/basic.py
@@ -531,16 +531,6 @@ def _c_array(ctype: type, values: List[Any]) -> ctypes.Array:
     return (ctype * len(values))(*values)  # type: ignore[operator]
 
 
-def _json_default_with_numpy(obj: Any) -> Any:
-    """Convert numpy classes to JSON serializable objects."""
-    if isinstance(obj, (np.integer, np.floating, np.bool_)):
-        return obj.item()
-    elif isinstance(obj, np.ndarray):
-        return obj.tolist()
-    else:
-        return obj
-
-
 def _to_string(x: Union[int, float, str, List]) -> str:
     if isinstance(x, list):
         val_list = ",".join(str(val) for val in x)
@@ -876,7 +866,7 @@ def _dump_pandas_categorical(
     pandas_categorical: Optional[List[List]],
     file_name: Optional[Union[str, Path]] = None,
 ) -> str:
-    categorical_json = json.dumps(pandas_categorical, default=_json_default_with_numpy)
+    categorical_json = json.dumps(pandas_categorical)
     pandas_str = f"\npandas_categorical:{categorical_json}\n"
     if file_name is not None:
         with open(file_name, "a") as f:
@@ -4700,12 +4690,7 @@ def dump_model(
                 )
             )
         ret = json.loads(string_buffer.value.decode("utf-8"), object_hook=object_hook)
-        ret["pandas_categorical"] = json.loads(
-            json.dumps(
-                self.pandas_categorical,
-                default=_json_default_with_numpy,
-            )
-        )
+        ret["pandas_categorical"] = json.loads(json.dumps(self.pandas_categorical))
         return ret
 
     def predict(
diff --git a/tests/python_package_test/test_basic.py b/tests/python_package_test/test_basic.py
index 108f37da0bee..5bd027f17179 100644
--- a/tests/python_package_test/test_basic.py
+++ b/tests/python_package_test/test_basic.py
@@ -880,6 +880,48 @@ def test_categorical_code_conversion_doesnt_modify_original_data(feature_name, c
     np.testing.assert_equal(codes, data[:, 0])
 
 
+def test_pandas_categorical_json_serialization_works(rng):
+    # Some 'numpy' types aren't JSON-serializable by default, this checks
+    # that having them in pandas_categorical doesn't cause issues for JSON
+    # serialization of models.
+    pd = pytest.importorskip("pandas")
+    df = pd.DataFrame(
+        {
+            "np_float": pd.Categorical(np.array([1.1, 2.2, 3.3], dtype=np.float32)),
+            "np_bool": pd.Categorical(np.array([True, False, False], dtype=np.bool_)),
+            "np_int": pd.Categorical(np.array([1, 2, 3], dtype=np.int64)),
+        }
+    )
+
+    # confirm that the array dtypes also become the category dtypes
+    assert df["np_float"].dtype.categories.dtype in (np.float32, np.float64)
+    assert df["np_bool"].dtype.categories.dtype == np.bool_
+    assert df["np_int"].dtype.categories.dtype == np.int64
+
+    dtrain = lgb.Dataset(
+        df,
+        label=rng.random((df.shape[0],)),
+        categorical_feature=["np_float", "np_bool", "np_int"],
+    )
+    dtrain.construct()
+
+    bst = lgb.train(
+        params={
+            "objective": "regression",
+        },
+        train_set=dtrain,
+        num_boost_round=2,
+    )
+
+    # JSON serialization works
+    json_data = bst.dump_model()
+    assert json_data["pandas_categorical"] == [
+        [float(np.float32(1.1)), float(np.float32(2.2)), float(np.float32(3.3))],
+        [False, True],
+        [1, 2, 3],
+    ]
+
+
 @pytest.mark.parametrize("min_data_in_bin", [2, 10])
 def test_feature_num_bin(min_data_in_bin, rng):
     X = np.vstack(