diff --git a/onnxruntime/core/mlas/lib/sbconv_kernel_neon.cpp b/onnxruntime/core/mlas/lib/sbconv_kernel_neon.cpp
index f3a17f30be166..d513791c695f5 100644
--- a/onnxruntime/core/mlas/lib/sbconv_kernel_neon.cpp
+++ b/onnxruntime/core/mlas/lib/sbconv_kernel_neon.cpp
@@ -45,7 +45,12 @@ MlasConvPointwiseBf16KernelNeon(
     const bool ReluActivation = (KernelFlags & MLAS_CONV_KERNEL_FLAG_RELU_ACTIVATION) != 0;
 
     MLAS_BACKEND_KERNEL_SELECTOR_CONFIG mlas_backend_kernel_selector_config;
-    mlas_backend_kernel_selector_config.use_kleidiai = ((KernelFlags & MLAS_CONV_KERNEL_MLAS_ARM_USE_KLEIDIAI) != 0);
+
+    // TODO(hasesh): With the ARM KleidiAI team, study the impact of using the KleidiAI SBGEMM kernel for this convolution kernel
+    // and enable it if things look okay.
+    // Even if re-enabled, honor user override to disable KleidiAI usage if specified.
+    // That is, mlas_backend_kernel_selector_config.use_kleidiai = ((KernelFlags & MLAS_CONV_KERNEL_MLAS_ARM_USE_KLEIDIAI) != 0);
+    mlas_backend_kernel_selector_config.use_kleidiai = false;
 
     const size_t StrideWidthElements = StrideWidth / sizeof(float);
     const size_t InputStrideElements = InputStride / sizeof(float);
diff --git a/onnxruntime/core/mlas/lib/sbgemm.h b/onnxruntime/core/mlas/lib/sbgemm.h
index 559c8b48e78b6..35520c7a4226a 100644
--- a/onnxruntime/core/mlas/lib/sbgemm.h
+++ b/onnxruntime/core/mlas/lib/sbgemm.h
@@ -311,7 +311,7 @@ MlasSBGemmPackBSize(
     //
     // Compute the number of bytes required to hold the packed buffer.
     //
-#if defined(USE_KLEIDIAI) && !defined(MLAS_USE_ARM_NEON_NCHWC)
+#if defined(USE_KLEIDIAI)
     if ((!BackendKernelSelectorConfig || BackendKernelSelectorConfig->use_kleidiai) &&
         GetMlasPlatform().MlasSBGemmPackBSizeOverride != nullptr &&
         TransA == CBLAS_TRANSPOSE::CblasNoTrans &&
@@ -359,7 +359,7 @@ MlasSBGemmConvertPackB(
     const MLAS_BACKEND_KERNEL_SELECTOR_CONFIG* BackendKernelSelectorConfig
 )
 {
-#if defined(USE_KLEIDIAI) && !defined(MLAS_USE_ARM_NEON_NCHWC)
+#if defined(USE_KLEIDIAI)
     if ((!BackendKernelSelectorConfig || BackendKernelSelectorConfig->use_kleidiai) &&
         GetMlasPlatform().MlasSBGemmPackBOverride != nullptr &&
         TransA == CBLAS_TRANSPOSE::CblasNoTrans &&
@@ -393,7 +393,7 @@ MlasSBGemmBatch(
     const MLAS_BACKEND_KERNEL_SELECTOR_CONFIG* BackendKernelSelectorConfig
 )
 {
-#if defined(USE_KLEIDIAI) && !defined(MLAS_USE_ARM_NEON_NCHWC)
+#if defined(USE_KLEIDIAI)
     if ((!BackendKernelSelectorConfig || BackendKernelSelectorConfig->use_kleidiai) &&
         GetMlasPlatform().MlasSBGemmBatchOverride != nullptr &&
         TransA == CBLAS_TRANSPOSE::CblasNoTrans &&