@FractalEcho reminds me of the early attempts to deploy LLM detection tools in academia that ended up consistently flagging works of foreign students as generated because their linguistic expression was too different.