ChatGPT: When two years of academic work vanished with a single click (www.nature.com)

A professor at the University of Cologne experienced a significant loss of two years' worth of academic work when using ChatGPT after disabling the ‘data consent’ option. This action led to the irreversible deletion of carefully structured project folders and conversations he had relied upon for research, teaching materials, and grant applications. Despite being a paid subscriber, he found the lack of any warning or recovery option concerning, highlighting a critical gap in accountability and reliability for users employing AI tools in academic settings. This incident underscores the potential risks of integrating generative AI into research and teaching. While tools like ChatGPT offer convenience and flexibility, the absence of safeguards—such as confirmatory prompts for irreversible deletions and data redundancy—raises significant concerns. OpenAI acknowledged that prioritizing user privacy means deleted content cannot be recovered, advising users to maintain personal backups for professional work. This case serves as a stark reminder for the AI/ML community about the importance of developing robust safety measures and accountability in the rapidly evolving landscape of AI tools for education and research.