diff --git a/runtime/executor/method.cpp b/runtime/executor/method.cpp
index 4e219b07a66..cdda246887c 100644
--- a/runtime/executor/method.cpp
+++ b/runtime/executor/method.cpp
@@ -1540,6 +1540,25 @@ Error Method::execute() {
       "Cannot execute until method has been initialized.");
   ET_LOG(Debug, "Executing method: %s.", method_meta().name());
 
+  // Validate that outputs are set.
+  for (size_t i = 0; i < method_meta().num_outputs(); i++) {
+    auto& output = mutable_value(get_output_index(i));
+    if (!output.isTensor()) {
+      continue;
+    }
+
+    auto tensor_meta = method_meta().output_tensor_meta(i);
+    if (tensor_meta->is_memory_planned()) {
+      continue;
+    }
+
+    auto& t = output.toTensor();
+    if (t.const_data_ptr() == nullptr) {
+      ET_LOG(Error, "Output tensor at index %" ET_PRIsize_t " is not set.", i);
+      return Error::InvalidState;
+    }
+  }
+
   // Chains are executed sequentially today, but future async designs may
   // branch and run many in parallel or out of order.
   for (step_state_.chain_idx = 0; step_state_.chain_idx < n_chains_;
diff --git a/runtime/executor/test/method_test.cpp b/runtime/executor/test/method_test.cpp
index f597746e0fd..8cd5f26c8e6 100644
--- a/runtime/executor/test/method_test.cpp
+++ b/runtime/executor/test/method_test.cpp
@@ -390,3 +390,36 @@ TEST_F(MethodTest, OptionalTensorListDeserialization) {
   EXPECT_EQ(outputs.toTensor().size(2), 10);
 }
 */
+
+TEST_F(MethodTest, UnsetOutputTest) {
+  // Validate that methods with non-memory planned outputs return an error when
+  // the output data pointer is not set. It should not crash.
+
+  ManagedMemoryManager mmm(kDefaultNonConstMemBytes, kDefaultRuntimeMemBytes);
+  Result<Method> method = programs_["cat"]->load_method("forward", &mmm.get());
+  ASSERT_EQ(method.error(), Error::Ok);
+
+  constexpr int buffer_size = 8;
+  float buffer[buffer_size];
+  for (int i = 0; i < buffer_size; ++i) {
+    buffer[i] = 0.f;
+  }
+  int32_t sizes[2] = {2, 4};
+  uint8_t dim_order[2] = {0, 1};
+  int32_t strides[2] = {4, 1};
+  executorch::aten::TensorImpl impl(
+      executorch::aten::ScalarType::Float,
+      2,
+      sizes,
+      buffer,
+      dim_order,
+      strides);
+
+  auto input_err =
+      method->set_input(EValue(executorch::aten::Tensor(&impl)), 0);
+  ASSERT_EQ(input_err, Error::Ok);
+
+  // Execute the method once. It should return an error (not crash).
+  auto execute_error = method->execute();
+  ASSERT_EQ(execute_error, Error::InvalidState);
+}