OpenAI plans new tools to detect ChatGPT-generated writing

OpenAI has announced plans to release tools designed to detect writing generated by ChatGPT. This tool aims to identify students who may use the AI to complete their assignments. The company is exploring text watermarking methods, which have shown high accuracy and effectiveness against localized tampering, such as paraphrasing.

However, OpenAI is cautious about releasing the tool due to potential negative impacts on specific groups, including non-English speakers. This hesitancy follows the company's decision to shut down its previous AI text detector in July 2023, citing low accuracy rates.

In an updated blog post from May, OpenAI highlighted that while text watermarking is effective against localized tampering, it is less robust against globalized tampering. This includes methods like using translation systems, rewording with another generative model, or inserting special characters between words and then removing them.