diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 126098f5fa..9da87e3207 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -9,7 +9,7 @@ channels:
 dependencies:
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
-- cuda-python>=11.7.1,<12.0a0,<=11.8.3
+- cuda-python>=11.8.5,<12.0a0
 - cuda-version=11.8
 - cudatoolkit
 - cudf==25.2.*,>=0.0.0a0
diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
index bfa564b93c..7491de8057 100644
--- a/conda/environments/all_cuda-125_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -12,7 +12,7 @@ dependencies:
 - cuda-cudart-dev
 - cuda-nvcc
 - cuda-profiler-api
-- cuda-python>=12.0,<13.0a0,<=12.6.0
+- cuda-python>=12.6.2,<13.0a0
 - cuda-version=12.5
 - cudf==25.2.*,>=0.0.0a0
 - cupy>=12.0.0
diff --git a/conda/recipes/cuml/meta.yaml b/conda/recipes/cuml/meta.yaml
index a2418dad10..35157fb3cc 100644
--- a/conda/recipes/cuml/meta.yaml
+++ b/conda/recipes/cuml/meta.yaml
@@ -58,10 +58,10 @@ requirements:
     - cuda-version ={{ cuda_version }}
     {% if cuda_major == "11" %}
     - cudatoolkit
-    - cuda-python >=11.7.1,<12.0a0,<=11.8.3
+    - cuda-python >=11.8.5,<12.0a0
     {% else %}
     - cuda-cudart-dev
-    - cuda-python >=12.0,<13.0a0,<=12.6.0
+    - cuda-python >=12.6.2,<13.0a0
     {% endif %}
     - cudf ={{ minor_version }}
     - cython >=3.0.0
@@ -77,10 +77,10 @@ requirements:
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
     {% if cuda_major == "11" %}
     - cudatoolkit
-    - cuda-python >=11.7.1,<12.0a0,<=11.8.3
+    - cuda-python >=11.8.5,<12.0a0
     {% else %}
     - cuda-cudart
-    - cuda-python >=12.0,<13.0a0,<=12.6.0
+    - cuda-python >=12.6.2,<13.0a0
     {% endif %}
     - cudf ={{ minor_version }}
     - cupy >=12.0.0
diff --git a/dependencies.yaml b/dependencies.yaml
index 83833e6f01..1110eba5dc 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -198,11 +198,11 @@ dependencies:
           - matrix:
               cuda: "12.*"
             packages:
-              - cuda-python>=12.0,<13.0a0,<=12.6.0
+              - cuda-python>=12.6.2,<13.0a0
           - matrix:
               cuda: "11.*"
             packages:
-              - cuda-python>=11.7.1,<12.0a0,<=11.8.3
+              - cuda-python>=11.8.5,<12.0a0
           - matrix:
             packages:
               - cuda-python
diff --git a/python/cuml/cuml/dask/neighbors/kneighbors_classifier.py b/python/cuml/cuml/dask/neighbors/kneighbors_classifier.py
index 18babd75c0..fa0e1e59d9 100644
--- a/python/cuml/cuml/dask/neighbors/kneighbors_classifier.py
+++ b/python/cuml/cuml/dask/neighbors/kneighbors_classifier.py
@@ -114,8 +114,6 @@ def fit(self, X, y):
                 # Dask-expr does not support numerical column names
                 # See: https://github.com/dask/dask-expr/issues/1015
                 _y = y
-                if hasattr(y, "to_legacy_dataframe"):
-                    _y = y.to_legacy_dataframe()
                 n_targets = len(_y.columns)
                 for i in range(n_targets):
                     uniq_labels.append(_y.iloc[:, i].unique())
diff --git a/python/cuml/cuml/tests/dask/test_dask_arr_utils.py b/python/cuml/cuml/tests/dask/test_dask_arr_utils.py
index 17657346d1..f009c04e9b 100644
--- a/python/cuml/cuml/tests/dask/test_dask_arr_utils.py
+++ b/python/cuml/cuml/tests/dask/test_dask_arr_utils.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2023, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -50,6 +50,10 @@ def test_to_sparse_dask_array(input_type, nrows, ncols, client):
 
     a = cupyx.scipy.sparse.random(nrows, ncols, format="csr", dtype=cp.float32)
     if input_type == "dask_dataframe":
+        pytest.xfail(
+            reason="Dask nightlies break task fusing for this, "
+            "issue https://github.com/rapidsai/cuml/issues/6169"
+        )
         df = cudf.DataFrame(a.todense())
         inp = dask_cudf.from_cudf(df, npartitions=2)
     elif input_type == "dask_array":
diff --git a/python/cuml/cuml/tests/dask/test_dask_logistic_regression.py b/python/cuml/cuml/tests/dask/test_dask_logistic_regression.py
index e73586498f..94ebecf6b0 100644
--- a/python/cuml/cuml/tests/dask/test_dask_logistic_regression.py
+++ b/python/cuml/cuml/tests/dask/test_dask_logistic_regression.py
@@ -80,9 +80,13 @@ def cal_chunks(dataset, n_partitions):
     X_da = da.from_array(X_train, chunks=(target_chunk_sizes, -1))
     y_da = da.from_array(y_train, chunks=target_chunk_sizes)
 
-    X_da, y_da = dask_utils.persist_across_workers(
-        c, [X_da, y_da], workers=workers
-    )
+    # todo (dgd): Dask nightly packages break persisting
+    # sparse arrays before using them.
+    # https://github.com/rapidsai/cuml/issues/6168
+
+    # X_da, y_da = dask_utils.persist_across_workers(
+    #     c, [X_da, y_da], workers=workers
+    # )
     return X_da, y_da