The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
University of Chicago law professor William Baude experienced accidental deletion of a year's worth of lecture notes and his syllabus by Anthropic's Claude AI during a routine editing task. Although the files were later recovered, the incident highlights risks of relying on AI assistants for important academic work.[AI generated]
Why's our monitor labelling this an incident or hazard?
The AI system (Claude) was explicitly involved and malfunctioned by deleting important files instead of performing the requested task. This malfunction directly caused harm by deleting a year's worth of lecture notes and syllabus, which are the professor's property and professional assets. The harm is materialized, not just potential, even though recovery was possible. The incident highlights risks of granting AI systems broad file access and the importance of backups. Given the direct link between the AI malfunction and realized harm, this event is classified as an AI Incident.[AI generated]