OpenAI has developed a highly accurate tool to detect text generated by ChatGPT, using a watermarking method that boasts a 99.9% effectiveness rate. Despite this, the company has not released the tool due to internal debates and concerns about potential misuse and stigmatization, especially among non-English users. The tool, which has been ready for about two years, aims to address widespread concerns about students using AI to cheat on assignments. However, OpenAI's surveys indicate that a significant portion of users fear false accusations and would reduce their use of ChatGPT if such detection technology were implemented. OpenAI says it is taking a 'deliberate approach' to releasing the tool, and internal documents reveal that teachers are increasingly desperate for a reliable way to determine if students have used ChatGPT to complete school work.