It's an alignment problem in the sense that it demonstrates the principle that today's AI systems cannot be trusted to reliably work towards the goals of their users. A small-scale alignment failure and a large-scale alignment failure are the same fundamental type of failure. Typically, large disasters come after smaller disasters which foreshadowed the disaster mechanism, but weren't taken seriously.
An AI which is mistaken about the goal you give it, or how to go about achieving it, will behave in a de facto malicious manner, as this incident illustrates.
