diff --git a/src/plugins/intel_npu/src/plugin/npuw/llm_compiled_model.cpp b/src/plugins/intel_npu/src/plugin/npuw/llm_compiled_model.cpp
index e18b098969eb79..260a1c444284cb 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/llm_compiled_model.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/llm_compiled_model.cpp
@@ -30,13 +30,13 @@ std::shared_ptr<ov::Model> redirect_new_kv_to_output(const std::shared_ptr<ov::M
 std::shared_ptr<ov::Model> cvt_kvcache_to_fp16(const std::shared_ptr<ov::Model>& model) {
     ov::preprocess::PrePostProcessor ppp(model);
 
-    for (auto tensor : model->inputs()) {
+    for (const auto& tensor : model->inputs()) {
         if (tensor.get_any_name().find("past_key") != std::string::npos) {
             ppp.input(tensor.get_any_name()).tensor().set_element_type(ov::element::Type_t::f16);
         }
     }
 
-    for (auto tensor : model->outputs()) {
+    for (const auto& tensor : model->outputs()) {
         if (tensor.get_any_name().find("present") != std::string::npos) {
             ppp.output(tensor.get_any_name()).tensor().set_element_type(ov::element::Type_t::f16);
         }
@@ -55,7 +55,7 @@ void reshape_to_static(std::shared_ptr<ov::Model> model,
                        const uint32_t kvcache_size,
                        const KVAxesPosition& kv_axes_position) {
     std::map<std::string, ov::PartialShape> new_shapes;
-    for (auto input : model->inputs()) {
+    for (const auto& input : model->inputs()) {
         const auto& input_name = input.get_any_name();
         ov::PartialShape new_shape;
         if (input_name.find("input_ids") != std::string::npos) {
@@ -275,7 +275,7 @@ ov::npuw::LLMCompiledModel::LLMCompiledModel(const std::shared_ptr<ov::Model>& m
 
     auto npudesc = extract_npu_descriptor(plugin);
 
-    ov::AnyMap properties_copy = other_props;
+    ov::AnyMap properties_copy = std::move(other_props);
     auto prefill_config = get_default_prefill_config(model, npudesc);
     // NB: GENERATE_HINT is only applicable for default generate config!
     const ::intel_npu::npuw::llm::GenerateHint generate_hint = m_cfg.get<::intel_npu::NPUW_LLM_GENERATE_HINT>();
diff --git a/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp b/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
index a8c90884d3d926..a73478c0cab5d2 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/llm_infer_request.cpp
@@ -36,17 +36,17 @@ ov::npuw::LLMInferRequest::LLMInferRequest(const std::shared_ptr<ov::npuw::LLMCo
     m_kvcache_request = compiled_model->m_kvcache_compiled->create_infer_request();
     m_prefill_request = compiled_model->m_prefill_compiled->create_infer_request();
 
-    for (auto input_port : m_prefill_request->get_compiled_model()->inputs()) {
+    for (const auto& input_port : m_prefill_request->get_compiled_model()->inputs()) {
         m_prefill_in_ports.emplace(input_port.get_any_name(), input_port);
     }
-    for (auto output_port : m_prefill_request->get_compiled_model()->outputs()) {
+    for (const auto& output_port : m_prefill_request->get_compiled_model()->outputs()) {
         m_prefill_out_ports.emplace(output_port.get_any_name(), output_port);
     }
 
-    for (auto input_port : m_kvcache_request->get_compiled_model()->inputs()) {
+    for (const auto& input_port : m_kvcache_request->get_compiled_model()->inputs()) {
         m_kvcache_in_ports.emplace(input_port.get_any_name(), input_port);
     }
-    for (auto output_port : m_kvcache_request->get_compiled_model()->outputs()) {
+    for (const auto& output_port : m_kvcache_request->get_compiled_model()->outputs()) {
         m_kvcache_out_ports.emplace(output_port.get_any_name(), output_port);
     }
 }
diff --git a/src/plugins/intel_npu/src/plugin/npuw/weights_bank.cpp b/src/plugins/intel_npu/src/plugin/npuw/weights_bank.cpp
index 5ff064e7629759..ddc8f10ccf034e 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/weights_bank.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/weights_bank.cpp
@@ -132,10 +132,15 @@ ov::Tensor Bank::eval_and_alloc(const LazyTensor& tensor,
 
 bool Bank::is_remote(const LazyTensor& tensor) const {
     // FIXME: make generic
+    std::lock_guard<std::mutex> guard(m_mutex);
+
     auto npu_bank = m_device_banks.find("NPU");
-    if (npu_bank != m_device_banks.end() && npu_bank->second.storage.find(tensor) != npu_bank->second.storage.end()) {
-        // Found in NPU bank so considered remote (utterly wrong for the generic case)
-        return true;
+    if (npu_bank != m_device_banks.end()) {
+        std::lock_guard<std::mutex> dev_guard(npu_bank->second.mutex);
+        if (npu_bank->second.storage.find(tensor) != npu_bank->second.storage.end()) {
+            // Found in NPU bank so considered remote (utterly wrong for the generic case)
+            return true;
+        }
     }
     return false;
 }
diff --git a/src/plugins/intel_npu/src/plugin/npuw/weights_bank.hpp b/src/plugins/intel_npu/src/plugin/npuw/weights_bank.hpp
index 491e962a58b438..f2ca0436607fd4 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/weights_bank.hpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/weights_bank.hpp
@@ -38,13 +38,13 @@ class Bank {
     // Bank for specified device and their allocated memory
     struct DeviceBank {
         std::unordered_map<LazyTensor, ov::Tensor, LazyTensor::Hash> storage;
-        std::mutex mutex;
+        mutable std::mutex mutex;
     };
     std::unordered_map<std::string, DeviceBank> m_device_banks;
 
     ov::Tensor eval_and_alloc(const LazyTensor& tensor, DeviceBank& dbank, const std::string& device);
 
-    std::mutex m_mutex;
+    mutable std::mutex m_mutex;
     std::shared_ptr<const ov::ICore> m_core = nullptr;
     std::string m_alloc_device;
 };