diff --git a/tests/fuzz/src/read_model-fuzzer.cc b/tests/fuzz/src/read_model-fuzzer.cc
new file mode 100644
index 00000000000..16b40dce084
--- /dev/null
+++ b/tests/fuzz/src/read_model-fuzzer.cc
@@ -0,0 +1,39 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "openvino/openvino.hpp"
+
+#define COUNT_OF(A) (sizeof(A) / sizeof(A[0]))
+const char kSplitSequence[] = {'F', 'U', 'Z', 'Z', '_', 'N', 'E', 'X', 'T', '_', 'F', 'I', 'E', 'L', 'D'};
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size) {
+    size_t split_counter = 0;
+    size_t split[1] = {0};
+    if (size < sizeof(kSplitSequence)) return 0;  // we at least expect one separator
+    for (size_t i = 0; i < size - sizeof(kSplitSequence); i++)
+        if (0 == memcmp(data + i, kSplitSequence, sizeof(kSplitSequence))) {
+            split[split_counter++] = i;
+            if (COUNT_OF(split) <= split_counter) break;
+        }
+    if (COUNT_OF(split) != split_counter) return 0;  // not enough splits
+
+    // isolate xml data
+    size_t net_size = split[0];
+    std::string net((const char*)data, net_size);
+    size -= net_size + sizeof(kSplitSequence);
+    data += net_size + sizeof(kSplitSequence);
+
+    // isolate weights data
+    ov::runtime::Tensor weights(ov::element::u8, {static_cast<size_t>(size)}, (void*)data);
+
+    // read xml and set weights
+    try {
+        ov::runtime::Core core;
+        auto model = core.read_model(net, weights);
+    } catch (const std::exception&) {
+        return 0;  // fail gracefully on expected exceptions
+    }
+
+    return 0;
+}
\ No newline at end of file