AI detector mislabels non-native English speaker essays as bot-generated

An AI detector has come under scrutiny after incorrectly labeling essays written by non-native English speakers as being generated by a bot. Researchers from Stanford University discovered that popular GPT detectors, which power chatbots such as ChatGPT, misidentified over half of the essays written by non-native English speakers as AI-generated. In contrast, the same systems correctly identified around 20% of essays written by native English-speaking eighth graders as being produced by AI. The researchers warned that this misclassification could have serious consequences for non-native English speakers, who could be falsely accused of using chatbots for their work. The study highlights the challenges and biases inherent in AI systems, and questions the effectiveness of AI-detection software in combatting misinformation.