OpenAI Acknowledges ChatGPT’s Inability to Detect AI-Generated Content, Offers Guidance to Educators

Sep 4, 2023

OpenAI, the artificial intelligence research lab, has issued a crucial advisory to educators, underlining that their AI chatbot, ChatGPT, is not equipped to identify AI-generated content. The organization has also provided a series of tips to help teachers recognize copied answers.

In a section of frequently asked questions tailored for educators, OpenAI admitted that while there exist tools designed to distinguish between AI and human-generated content, their reliability is questionable. The organization has alerted teachers about this issue as students often resort to AI tools to complete their assignments and homework.

OpenAI stated in its FAQ section that “Despite several organizations, including ourselves, having released tools claiming to detect AI-generated content, none have proven to be fully reliable in distinguishing between AI and human-created content.”

Regarding its own chatbot, OpenAI disclosed that ChatGPT lacks the capability to identify what content might be AI-generated. The organization clarified that even if users attempt to inquire about this from the chatbot using various prompts, the responses will be arbitrary and devoid of factual basis.

OpenAI further explained that during their attempts to train AI tools to recognize AI-generated content, the tools occasionally mislabeled classic human-authored content, such as works by Shakespeare or the Declaration of Independence, as AI-generated. OpenAI also noted that “Even if these tools could accurately identify AI-generated content (which they cannot yet), students can make minor edits to avoid detection.”

To address this challenge, OpenAI has put forth a set of guidelines for educators. To determine whether students have plagiarized their work from ChatGPT, OpenAI recommends teachers ask unique, content-specific questions to gauge the relevance of the response and the student’s understanding of the assignment. However, the organization emphasizes the need for educators to verify the information for potential biases.

OpenAI added, “Ensuring that students share their interactions with the model holds them accountable for their use of AI in their work. This allows educators to confirm that students are using the tool responsibly and meaningfully, rather than simply copying answers.”

According to OpenAI’s guidelines, ChatGPT users must be at least 13 years old. Users aged between 13 and 18 years are required to obtain parental or guardian permission to use ChatGPT.