OpenAI dumps its AI detection tool six months after launch

Less than six months after OpenAI introduced its AI detection tool, aimed at assisting teachers and professionals in identifying AI-generated work, the company has decided to pull the feature from its offerings.

The decision to shut down the tool was disclosed through an update to the original company blog post that had initially announced the feature.

According to OpenAI, the primary reason for discontinuing the tool was its "low rate of accuracy".

The company acknowledged the need for improvement and stated that they are actively working on incorporating user feedback while researching more effective provenance techniques for text. Additionally, OpenAI expressed its commitment to aiding users in determining if audio or visual content has been AI-generated.

The AI detection tool was first unveiled by OpenAI in February with the purpose of enabling users to determine whether an essay or text was authored by a human or AI. Focused on AI-generated English text, the tool operated on a machine learning system, which analysed the input and assigned it to various categories. By pasting a body of text, such as a school essay, into the tool, users received one of five possible outcomes, ranging from "likely generated by AI" to "very unlikely". However, even upon its launch, OpenAI openly admitted that the tool was not perfect and advised users to interpret the results with caution.