Anthropic has one of the most advanced LLMs, yet they still use regex to detect/block offensive words (per leaked source code of Claude Code). Not every problem needs to be solved with weapons-grade AI.