diff --git a/integration_tests/src/main/python/orc_write_test.py b/integration_tests/src/main/python/orc_write_test.py
index 3f70a7afa6a..2d58d198903 100644
--- a/integration_tests/src/main/python/orc_write_test.py
+++ b/integration_tests/src/main/python/orc_write_test.py
@@ -15,6 +15,7 @@
 import pytest
 
 from asserts import assert_gpu_and_cpu_writes_are_equal_collect, assert_gpu_fallback_write
+from spark_session import is_databricks104_or_later
 from datetime import date, datetime, timezone
 from data_gen import *
 from marks import *
@@ -172,6 +173,7 @@ def create_empty_df(spark, path):
 @pytest.mark.parametrize("provider", ["", "hadoop"])
 @pytest.mark.parametrize("encrypt", ["", "pii:a"])
 @pytest.mark.parametrize("mask", ["", "sha256:a"])
+@pytest.mark.skipif(is_databricks104_or_later(), reason="The test will fail on Databricks10.4 because `HadoopShimsPre2_3$NullKeyProvider` is loaded")
 def test_orc_write_encryption_fallback(spark_tmp_path, spark_tmp_table_factory, path, provider, encrypt, mask):
     def write_func(spark, write_path):
         writer = unary_op_df(spark, gen).coalesce(1).write