Lack of knowledge of a language is much less of an impediment to making changes than it used to be. An LLM can create syntactically correct code that the person operating the LLM cannot or won't bother to understand or verify. Even the best LLMs get a LOT wrong. 100% this is what they're doing.