diff --git a/level_zero/core/source/cmdqueue/cmdqueue_hw.h b/level_zero/core/source/cmdqueue/cmdqueue_hw.h
index fd97df19e2..e2acc5c3d5 100644
--- a/level_zero/core/source/cmdqueue/cmdqueue_hw.h
+++ b/level_zero/core/source/cmdqueue/cmdqueue_hw.h
@@ -156,7 +156,7 @@ struct CommandQueueHw : public CommandQueueImp {
     inline void assignCsrTaskCountToFenceIfAvailable(ze_fence_handle_t hFence);
     inline void dispatchTaskCountPostSyncRegular(bool isDispatchTaskCountPostSyncRequired, NEO::LinearStream &commandStream);
     inline void dispatchTaskCountPostSyncByMiFlushDw(bool isDispatchTaskCountPostSyncRequired, NEO::LinearStream &commandStream);
-    inline NEO::SubmissionStatus prepareAndSubmitBatchBuffer(CommandListExecutionContext &ctx, NEO::LinearStream &innerCommandStream);
+    NEO::SubmissionStatus prepareAndSubmitBatchBuffer(CommandListExecutionContext &ctx, NEO::LinearStream &innerCommandStream);
     inline void updateTaskCountAndPostSync(bool isDispatchTaskCountPostSyncRequired);
     inline ze_result_t waitForCommandQueueCompletionAndCleanHeapContainer();
     inline ze_result_t handleSubmissionAndCompletionResults(NEO::SubmissionStatus submitRet, ze_result_t completionRet);
diff --git a/level_zero/core/source/cmdqueue/cmdqueue_hw.inl b/level_zero/core/source/cmdqueue/cmdqueue_hw.inl
index 9a5112df37..eaa6004222 100644
--- a/level_zero/core/source/cmdqueue/cmdqueue_hw.inl
+++ b/level_zero/core/source/cmdqueue/cmdqueue_hw.inl
@@ -1017,7 +1017,15 @@ NEO::SubmissionStatus CommandQueueHw<gfxCoreFamily>::prepareAndSubmitBatchBuffer
         *(MI_BATCH_BUFFER_END *)buffer = GfxFamily::cmdInitBatchBufferEnd;
     }
 
-    if (this->alignedChildStreamPadding) {
+    if (ctx.isNEODebuggerActive(this->device) || NEO::DebugManager.flags.EnableSWTags.get()) {
+        auto leftoverSpace = outerCommandStream.getUsed() - innerCommandStream.getUsed();
+        leftoverSpace -= ptrDiff(innerCommandStream.getCpuBase(), outerCommandStream.getCpuBase());
+
+        if (leftoverSpace > 0) {
+            auto memory = innerCommandStream.getSpace(leftoverSpace);
+            memset(memory, 0, leftoverSpace);
+        }
+    } else if (this->alignedChildStreamPadding) {
         void *paddingPtr = innerCommandStream.getSpace(this->alignedChildStreamPadding);
         memset(paddingPtr, 0, this->alignedChildStreamPadding);
     }
diff --git a/level_zero/core/test/unit_tests/mocks/mock_cmdqueue.h b/level_zero/core/test/unit_tests/mocks/mock_cmdqueue.h
index 5778c2b076..2c070f73fc 100644
--- a/level_zero/core/test/unit_tests/mocks/mock_cmdqueue.h
+++ b/level_zero/core/test/unit_tests/mocks/mock_cmdqueue.h
@@ -61,6 +61,7 @@ template <GFXCORE_FAMILY gfxCoreFamily>
 struct MockCommandQueueHw : public L0::CommandQueueHw<gfxCoreFamily> {
     using BaseClass = ::L0::CommandQueueHw<gfxCoreFamily>;
     using BaseClass::commandStream;
+    using BaseClass::prepareAndSubmitBatchBuffer;
     using BaseClass::printfFunctionContainer;
     using L0::CommandQueue::activeSubDevices;
     using L0::CommandQueue::internalUsage;
@@ -68,6 +69,7 @@ struct MockCommandQueueHw : public L0::CommandQueueHw<gfxCoreFamily> {
     using L0::CommandQueue::partitionCount;
     using L0::CommandQueue::preemptionCmdSyncProgramming;
     using L0::CommandQueueImp::csr;
+    using typename BaseClass::CommandListExecutionContext;
 
     MockCommandQueueHw(L0::Device *device, NEO::CommandStreamReceiver *csr, const ze_command_queue_desc_t *desc) : L0::CommandQueueHw<gfxCoreFamily>(device, csr, desc) {
     }
diff --git a/level_zero/core/test/unit_tests/sources/cmdqueue/test_cmdqueue_1.cpp b/level_zero/core/test/unit_tests/sources/cmdqueue/test_cmdqueue_1.cpp
index eaf8fe00b0..597bfe9b7d 100644
--- a/level_zero/core/test/unit_tests/sources/cmdqueue/test_cmdqueue_1.cpp
+++ b/level_zero/core/test/unit_tests/sources/cmdqueue/test_cmdqueue_1.cpp
@@ -377,6 +377,39 @@ HWTEST2_F(CommandQueueCreate, givenGpuHangInReservingLinearStreamWhenExecutingCo
     EXPECT_EQ(ZE_RESULT_ERROR_DEVICE_LOST, result);
 }
 
+HWTEST2_F(CommandQueueCreate, givenSwTagsEnabledWhenPrepareAndSubmitBatchBufferThenLeftoverIsZeroed, IsAtLeastSkl) {
+    DebugManagerStateRestore restorer;
+    NEO::DebugManager.flags.EnableSWTags.set(1);
+    const ze_command_queue_desc_t desc = {};
+    auto commandQueue = new MockCommandQueueHw<gfxCoreFamily>(device, neoDevice->getDefaultEngine().commandStreamReceiver, &desc);
+    commandQueue->initialize(false, false);
+    ze_result_t returnValue;
+    auto commandList = std::unique_ptr<CommandList>(whiteboxCast(CommandList::create(productFamily, device, NEO::EngineGroupType::RenderCompute, 0u, returnValue)));
+    ASSERT_NE(nullptr, commandList);
+    auto &commandStream = commandQueue->commandStream;
+
+    auto estimatedSize = 4096u;
+    NEO::LinearStream linearStream(commandStream->getSpace(estimatedSize), estimatedSize);
+    // fill with random data
+    memset(commandStream->getCpuBase(), 0xD, estimatedSize);
+    typename MockCommandQueueHw<gfxCoreFamily>::CommandListExecutionContext ctx{};
+
+    commandQueue->prepareAndSubmitBatchBuffer(ctx, linearStream);
+
+    // MI_BATCH_BUFFER END will be added during prepareAndSubmitBatchBuffer
+    auto offsetInBytes = sizeof(typename FamilyType::MI_BATCH_BUFFER_END);
+    auto isLeftoverZeroed = true;
+    for (auto i = offsetInBytes; i < estimatedSize; i++) {
+        uint8_t *data = reinterpret_cast<uint8_t *>(commandStream->getCpuBase());
+        if (data[i] != 0) {
+            isLeftoverZeroed = false;
+            break;
+        }
+    }
+    EXPECT_TRUE(isLeftoverZeroed);
+    commandQueue->destroy();
+}
+
 template <GFXCORE_FAMILY gfxCoreFamily>
 struct MockCommandQueueHwEstimateSizeTest : public MockCommandQueueHw<gfxCoreFamily> {
 
diff --git a/level_zero/core/test/unit_tests/sources/debugger/test_l0_debugger_sba_tracking.cpp b/level_zero/core/test/unit_tests/sources/debugger/test_l0_debugger_sba_tracking.cpp
index 35388466f6..7bd5923652 100644
--- a/level_zero/core/test/unit_tests/sources/debugger/test_l0_debugger_sba_tracking.cpp
+++ b/level_zero/core/test/unit_tests/sources/debugger/test_l0_debugger_sba_tracking.cpp
@@ -330,6 +330,35 @@ HWTEST2_F(L0DebuggerTest, givenDebuggingEnabledWhenCommandListIsExecutedThenSbaB
     commandList->destroy();
 }
 
+HWTEST2_F(L0DebuggerTest, givenDebugerEnabledWhenPrepareAndSubmitBatchBufferThenLeftoverIsZeroed, Gen12Plus) {
+    ze_command_queue_desc_t queueDesc = {};
+    std::unique_ptr<MockCommandQueueHw<gfxCoreFamily>, Deleter> commandQueue(new MockCommandQueueHw<gfxCoreFamily>(device, neoDevice->getDefaultEngine().commandStreamReceiver, &queueDesc));
+    commandQueue->initialize(false, false);
+
+    auto &commandStream = commandQueue->commandStream;
+    auto estimatedSize = 4096u;
+    NEO::LinearStream linearStream(commandStream->getSpace(estimatedSize), estimatedSize);
+    // fill with random data
+    memset(commandStream->getCpuBase(), 0xD, estimatedSize);
+
+    typename MockCommandQueueHw<gfxCoreFamily>::CommandListExecutionContext ctx{};
+    ctx.isDebugEnabled = true;
+
+    commandQueue->prepareAndSubmitBatchBuffer(ctx, linearStream);
+
+    // MI_BATCH_BUFFER END is added during prepareAndSubmitBatchBuffer
+    auto offsetInBytes = sizeof(typename FamilyType::MI_BATCH_BUFFER_END);
+    auto isLeftoverZeroed = true;
+    for (auto i = offsetInBytes; i < estimatedSize; i++) {
+        uint8_t *data = reinterpret_cast<uint8_t *>(commandStream->getCpuBase());
+        if (data[i] != 0) {
+            isLeftoverZeroed = false;
+            break;
+        }
+    }
+    EXPECT_TRUE(isLeftoverZeroed);
+}
+
 INSTANTIATE_TEST_CASE_P(SBAModesForDebugger, L0DebuggerParameterizedTests, ::testing::Values(0, 1));
 
 struct L0DebuggerSingleAddressSpace : public Test<L0DebuggerHwFixture> {