diff --git a/tests/distributed/test_weight_transfer.py b/tests/distributed/test_weight_transfer.py
index 04747e732..b370721b3 100644
--- a/tests/distributed/test_weight_transfer.py
+++ b/tests/distributed/test_weight_transfer.py
@@ -456,11 +456,13 @@ class TestIPCWeightTransferUpdateInfoValidation:
                 ipc_handles=ipc_handles,
             )
 
-    def test_valid_update_info_from_pickled(self):
+    def test_valid_update_info_from_pickled(self, monkeypatch):
         """Test creating IPCWeightTransferUpdateInfo from pickled handles."""
         if torch.cuda.device_count() < 1:
             pytest.skip("Need at least 1 GPU for this test")
 
+        monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
+
         dummy_tensor = torch.ones(10, 10, device="cuda:0")
         ipc_handle = reduce_tensor(dummy_tensor)
         gpu_uuid = str(torch.cuda.get_device_properties(0).uuid)
@@ -477,6 +479,18 @@ class TestIPCWeightTransferUpdateInfoValidation:
         assert info.ipc_handles == ipc_handles
         assert info.ipc_handles_pickled is None
 
+    def test_pickled_requires_insecure_serialization_flag(self, monkeypatch):
+        """Test that pickled handles are rejected unless env flag is enabled."""
+        monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "0")
+
+        with pytest.raises(ValueError, match="VLLM_ALLOW_INSECURE_SERIALIZATION=1"):
+            IPCWeightTransferUpdateInfo(
+                names=[],
+                dtype_names=[],
+                shapes=[],
+                ipc_handles_pickled=base64.b64encode(pickle.dumps([])).decode("utf-8"),
+            )
+
     def test_both_handles_and_pickled_raises(self):
         """Test that providing both ipc_handles and ipc_handles_pickled raises."""
         if torch.cuda.device_count() < 1:
@@ -556,11 +570,13 @@ class TestIPCEngineParsing:
         assert update_info.shapes == [[100, 100], [50]]
         assert len(update_info.ipc_handles) == 2
 
-    def test_parse_update_info_pickled(self):
+    def test_parse_update_info_pickled(self, monkeypatch):
         """Test parsing update info with pickled IPC handles (HTTP path)."""
         if torch.cuda.device_count() < 1:
             pytest.skip("Need at least 1 GPU for this test")
 
+        monkeypatch.setenv("VLLM_ALLOW_INSECURE_SERIALIZATION", "1")
+
         config = WeightTransferConfig(backend="ipc")
         parallel_config = create_mock_parallel_config()
         engine = IPCWeightTransferEngine(config, parallel_config)
diff --git a/vllm/distributed/weight_transfer/ipc_engine.py b/vllm/distributed/weight_transfer/ipc_engine.py
index 2edbec625..85dd34553 100644
--- a/vllm/distributed/weight_transfer/ipc_engine.py
+++ b/vllm/distributed/weight_transfer/ipc_engine.py
@@ -12,6 +12,7 @@ import requests
 import torch
 from torch.multiprocessing.reductions import reduce_tensor
 
+from vllm import envs
 from vllm.config.parallel import ParallelConfig
 from vllm.config.weight_transfer import WeightTransferConfig
 from vllm.distributed.weight_transfer.base import (
@@ -74,6 +75,13 @@ class IPCWeightTransferUpdateInfo(WeightTransferUpdateInfo):
                 raise ValueError(
                     "Cannot specify both `ipc_handles` and `ipc_handles_pickled`"
                 )
+
+            if not envs.VLLM_ALLOW_INSECURE_SERIALIZATION:
+                raise ValueError(
+                    "Refusing to deserialize `ipc_handles_pickled` without "
+                    "VLLM_ALLOW_INSECURE_SERIALIZATION=1"
+                )
+
             self.ipc_handles = pickle.loads(base64.b64decode(self.ipc_handles_pickled))
             self.ipc_handles_pickled = None