
The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Researchers from Northwestern and the University of Chicago found that ChatGPT can write fake scientific abstracts convincing enough to fool human reviewers, with only 68% detected as AI-generated. While AI detectors identified all fakes, the incident highlights risks of academic deception and challenges in detecting AI-written content.[AI generated]
Why's our monitor labelling this an incident or hazard?
The event involves the use of an AI system (ChatGPT) to generate content that can mislead humans, specifically in the context of scientific abstracts. The misuse of AI-generated text in academic settings (e.g., students submitting AI-written work) constitutes a violation of academic integrity and potentially intellectual property rights, which falls under harm category (c) (violations of human rights or breach of obligations under applicable law, including intellectual property rights). The article describes realized harm in the form of deception and academic misconduct, not just potential harm. Therefore, this qualifies as an AI Incident. The discussion of detection methods and responses by OpenAI is complementary but secondary to the main incident of AI-generated deceptive content causing harm.[AI generated]