OpenAI appears to be holding back a new “highly accurate” tool capable of detecting content generated by ChatGPT over concerns that it could be tampered with or cause non-English users to avoid generating text with artificial intelligence models.