diff --git a/ggml-backend.c b/ggml-backend.c
index 56f96c3671ec1..8d66c5605de46 100644
--- a/ggml-backend.c
+++ b/ggml-backend.c
@@ -703,7 +703,7 @@ ggml_backend_t ggml_backend_cpu_init(void) {
 }
 
 bool ggml_backend_is_cpu(ggml_backend_t backend) {
-    return backend->iface.get_name == ggml_backend_cpu_name;
+    return backend && backend->iface.get_name == ggml_backend_cpu_name;
 }
 
 void ggml_backend_cpu_set_n_threads(ggml_backend_t backend_cpu, int n_threads) {
diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 032d7e98689e1..ee82f68d74e9a 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -10140,7 +10140,7 @@ ggml_backend_t ggml_backend_cuda_init(int device) {
 }
 
 bool ggml_backend_is_cuda(ggml_backend_t backend) {
-    return backend->iface.get_name == ggml_backend_cuda_name;
+    return backend && backend->iface.get_name == ggml_backend_cuda_name;
 }
 
 int ggml_backend_cuda_get_device_count() {
diff --git a/ggml-metal.m b/ggml-metal.m
index e3d76b9d3d332..00b377a3e1389 100644
--- a/ggml-metal.m
+++ b/ggml-metal.m
@@ -2749,7 +2749,7 @@ ggml_backend_t ggml_backend_metal_init(void) {
 }
 
 bool ggml_backend_is_metal(ggml_backend_t backend) {
-    return backend->iface.get_name == ggml_backend_metal_name;
+    return backend && backend->iface.get_name == ggml_backend_metal_name;
 }
 
 void ggml_backend_metal_set_n_cb(ggml_backend_t backend, int n_cb) {