diff --git a/src/plugins/intel_gpu/src/plugin/sync_infer_request.cpp b/src/plugins/intel_gpu/src/plugin/sync_infer_request.cpp
index 26771117e2e786..985336b801b9d3 100644
--- a/src/plugins/intel_gpu/src/plugin/sync_infer_request.cpp
+++ b/src/plugins/intel_gpu/src/plugin/sync_infer_request.cpp
@@ -592,6 +592,12 @@ void SyncInferRequest::allocate_input(const ov::Output<const ov::Node>& port, si
     auto element_type = port.get_element_type();
 
     m_user_inputs[input_idx] = { create_host_tensor(shape, element_type), TensorOwner::PLUGIN };
+    if (element_type == ov::element::string) {
+        // In case the element type is string and input data is an empty string,
+        // it produces the segmentation fault unless the each element of tensor.data is initialized.
+        auto data = m_user_inputs.at(input_idx).ptr->data<std::string>();
+        std::uninitialized_fill_n(data, m_user_inputs.at(input_idx).ptr->get_size(), std::string());
+    }
     ov::ISyncInferRequest::set_tensor(port, m_user_inputs.at(input_idx).ptr);
 }
 
diff --git a/tests/layer_tests/tensorflow_tests/test_tf_LookupTableSize.py b/tests/layer_tests/tensorflow_tests/test_tf_LookupTableSize.py
index e0050c245f1321..4cd5b05f3e86d4 100644
--- a/tests/layer_tests/tensorflow_tests/test_tf_LookupTableSize.py
+++ b/tests/layer_tests/tensorflow_tests/test_tf_LookupTableSize.py
@@ -69,8 +69,6 @@ def create_lookup_table_size_net(self, hash_table_type, keys_type, values_type,
     def test_lookup_table_size(self, hash_table_type, params, ie_device, precision, ir_version, temp_dir,
                                use_legacy_frontend):
         keys_type = params['keys_type']
-        if ie_device == 'GPU' and keys_type == str:
-            pytest.skip("148921: Segmentation fault on GPU")
         self._test(*self.create_lookup_table_size_net(hash_table_type=hash_table_type, **params),
                    ie_device, precision, ir_version, temp_dir=temp_dir,
                    use_legacy_frontend=use_legacy_frontend)