Google claims one of its AI models is the first of its kind to spot a memory safety vulnerability in the wild – specifically ...