diff --git a/econml/sklearn_extensions/linear_model.py b/econml/sklearn_extensions/linear_model.py
index 56091899c..2ccba55d6 100644
--- a/econml/sklearn_extensions/linear_model.py
+++ b/econml/sklearn_extensions/linear_model.py
@@ -437,7 +437,6 @@ def __init__(self, eps=1e-3, n_alphas=100, alphas=None, fit_intercept=True,
                 cv=cv, verbose=verbose, n_jobs=n_jobs, positive=positive,
                 random_state=random_state, selection=selection)
             self.n_alphas = n_alphas
-            self.alphas = alphas
         else:
             super().__init__(
                 eps=eps, n_alphas=n_alphas, alphas=alphas,
@@ -561,7 +560,6 @@ def __init__(self, eps=1e-3, n_alphas=100, alphas=None, fit_intercept=True,
                 cv=cv, verbose=verbose, n_jobs=n_jobs,
                 random_state=random_state, selection=selection)
             self.n_alphas = n_alphas
-            self.alphas = alphas
         else:
             super().__init__(
                 eps=eps, n_alphas=n_alphas, alphas=alphas,
diff --git a/econml/tests/test_linear_model.py b/econml/tests/test_linear_model.py
index bf28a5a5b..0608e2f93 100644
--- a/econml/tests/test_linear_model.py
+++ b/econml/tests/test_linear_model.py
@@ -296,6 +296,35 @@ def test_wrapper_attributes(self):
         assert wrapper.max_iter == 100
         assert wrapper.tol == 0.01
 
+    def test_default_alphas_fits_on_strict_sklearn(self):
+        # Regression test for #1032. WeightedLassoCV.__init__ used to overwrite
+        # self.alphas = None after the version-dispatch correctly translated
+        # the default n_alphas=100 into alphas=100 on the super().__init__()
+        # call. sklearn 1.9's strict param validation rejected the resulting
+        # None and SparseLinearDML.fit / DebiasedLasso.fit raised
+        # InvalidParameterError. With default args, the dispatched alphas value
+        # must survive to fit time.
+        from packaging.version import parse
+        import sklearn
+        if parse(sklearn.__version__) < parse("1.7"):
+            self.skipTest("dispatch only active on sklearn 1.7+")
+
+        for cls in (WeightedLassoCV, WeightedMultiTaskLassoCV):
+            est = cls()
+            assert est.alphas is not None, \
+                f"{cls.__name__}.alphas was clobbered to None by __init__ (#1032)"
+            assert est.n_alphas == 100, \
+                f"{cls.__name__}.n_alphas should be preserved at 100 (#1032)"
+
+        rng = np.random.default_rng(0)
+        n = 300
+        X = rng.normal(size=(n, 3))
+        y_1d = rng.normal(size=n)
+        y_2d = rng.normal(size=(n, 2))
+        # neither call should raise sklearn.InvalidParameterError
+        WeightedLassoCV(cv=3).fit(X, y_1d)
+        WeightedMultiTaskLassoCV(cv=3).fit(X, y_2d)
+
     #################
     # DebiasedLasso #
     #################
diff --git a/pyproject.toml b/pyproject.toml
index 615ce8672..1fd51b14b 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -27,7 +27,7 @@ dependencies = [
     # in addition to dependencies)
     "numba > 0.53.1",
     "scipy > 1.4.0",
-    "scikit-learn >= 1.0, < 1.9",
+    "scikit-learn >= 1.0, < 1.10",
     "sparse",
     "joblib >= 0.13.0",
     "statsmodels >= 0.10",