OpenAI Creates Tool to Detect Texts Written by ChatGPT

Hibapress

OpenAI has created a tool to detect texts written by ChatGPT in order to combat academic and school cheating. Although it is ready to be launched, its release has been delayed by a year due to internal debates and concerns about its impact.

OpenAI employees are divided: some want to maintain transparency, while others fear losing users. A survey shows that a third of regular users would be discouraged by this anti-cheat technology, reports the Wall Street Journal.

The tool works by inserting a detectable pattern into the generated text, which is 99.9% effective. But it can be circumvented by techniques like translation or adding emojis. It’s also difficult to decide who should have access to the tool.

OpenAI plans to provide this tool directly to educators or through companies that specialize in plagiarism detection.

The decision to release this tool will depend on its impact on users and the ecosystem. OpenAI wants to align this decision with its values ​​of transparency and accountability.