diff --git a/opm/simulators/linalg/bda/WellContributions.cpp b/opm/simulators/linalg/bda/WellContributions.cpp
index c284c9282..c09695132 100644
--- a/opm/simulators/linalg/bda/WellContributions.cpp
+++ b/opm/simulators/linalg/bda/WellContributions.cpp
@@ -62,10 +62,6 @@ WellContributions::~WellContributions()
     }
 #endif
 
-    if(num_std_wells > 0){
-        delete[] val_pointers;
-    }
-
 #if HAVE_OPENCL
     if(opencl_gpu){
         if(num_ms_wells > 0){
@@ -181,7 +177,7 @@ void WellContributions::addMatrix([[maybe_unused]] MatrixType type, [[maybe_unus
             if (num_std_wells_so_far == num_std_wells - 1) {
                 val_pointers[num_std_wells] = num_blocks;
                 events.resize(1);
-                queue->enqueueWriteBuffer(*d_val_pointers_ocl, CL_FALSE, 0, sizeof(unsigned int) * (num_std_wells + 1), val_pointers, nullptr, &events[0]);
+                queue->enqueueWriteBuffer(*d_val_pointers_ocl, CL_FALSE, 0, sizeof(unsigned int) * (num_std_wells + 1), val_pointers.data(), nullptr, &events[0]);
                 events[0].wait();
                 events.clear();
             }
@@ -227,7 +223,7 @@ void WellContributions::addNumBlocks(unsigned int numBlocks)
 void WellContributions::alloc()
 {
     if (num_std_wells > 0) {
-        val_pointers = new unsigned int[num_std_wells + 1];
+        val_pointers.resize(num_std_wells+1);
 
 #if HAVE_CUDA
         if(cuda_gpu){
diff --git a/opm/simulators/linalg/bda/WellContributions.cu b/opm/simulators/linalg/bda/WellContributions.cu
index 1ba663cf0..d374c5523 100644
--- a/opm/simulators/linalg/bda/WellContributions.cu
+++ b/opm/simulators/linalg/bda/WellContributions.cu
@@ -210,7 +210,7 @@ void WellContributions::addMatrixGpu(MatrixType type, int *colIndices, double *v
         val_pointers[num_std_wells_so_far] = num_blocks_so_far;
         if (num_std_wells_so_far == num_std_wells - 1) {
             val_pointers[num_std_wells] = num_blocks;
-            cudaMemcpy(d_val_pointers, val_pointers, sizeof(unsigned int) * (num_std_wells + 1), cudaMemcpyHostToDevice);
+            cudaMemcpy(d_val_pointers, val_pointers.data(), sizeof(unsigned int) * (num_std_wells + 1), cudaMemcpyHostToDevice);
         }
         break;
     default:
diff --git a/opm/simulators/linalg/bda/WellContributions.hpp b/opm/simulators/linalg/bda/WellContributions.hpp
index eec664b7c..c0735e715 100644
--- a/opm/simulators/linalg/bda/WellContributions.hpp
+++ b/opm/simulators/linalg/bda/WellContributions.hpp
@@ -86,7 +86,7 @@ private:
     unsigned int num_ms_wells = 0;           // number of MultisegmentWells in this object, must equal multisegments.size()
     unsigned int num_blocks_so_far = 0;      // keep track of where next data is written
     unsigned int num_std_wells_so_far = 0;   // keep track of where next data is written
-    unsigned int *val_pointers = nullptr;    // val_pointers[wellID] == index of first block for this well in Ccols and Bcols
+    std::vector<unsigned int> val_pointers;    // val_pointers[wellID] == index of first block for this well in Ccols and Bcols
 
     double *h_x = nullptr;
     double *h_y = nullptr;