Claude Deletes Chicago Professor's Lecture Files
A routine page-number edit wiped a syllabus and notes, recovered only through Dropbox, in a reported agent failure.
Updated on Nov. 6, 2026 in Artificial Intelligence
Live Poll
Would you trust an AI tool to edit your important work files?
Claude, Anthropic's AI model, accidentally deleted University of Chicago law professor William Baude's Fall 2026 syllabus and all top-level Class N 2023 lecture notes while it was updating page numbers in an old set of notes. The deleted files were later recovered via Dropbox. This is a reported incident involving a deployed AI tool, not a new product announcement.
Why it matters
The episode illustrates how an AI agent given routine editing access can cause destructive, unintended changes to working files. It fits a wider pattern of reported agent mishaps in which automated tools have removed production data.
The incident involved a single page-number check that surfaced one incorrect citation, after which the agent removed files it had been tasked only with editing. A separate reported case saw a Claude agent wipe an entire company database in 9 seconds, showing how quickly an agent's file or database operations can execute.
The players
Claude
Anthropic's AI model family, here operating as an agent performing file edits on a user's documents.
William Baude
Harry Kalven, Jr. Professor of Law at the University of Chicago Law School, whose files were affected.
University of Chicago
Private research university whose law school employs Professor William Baude.
Dropbox
Cloud file-storage and sync service that held the copy from which the deleted files were recovered.
The details
Baude tasked Claude with updating page numbers in an old set of lecture notes. When he returned from lunch he found a note from Claude reporting that it had deleted the files by accident, including the Fall 2026 final syllabus and all top-level Class N 2023 lecture notes. The files were later recovered through Dropbox, which kept a copy from which they could be restored.
Timeline
The Chicago deletion was reported without a specific date in the source material.
In April, a Claude agent deleted an entire company database in 9 seconds.
In October, an AI agent reportedly deleted an entire TV show's edit weeks before release.
The Tech Race
Claude's file-handling errors place it alongside other AI agents that have run destructive commands during routine work, including an agent reported to have deleted a TV show's edit weeks before release in October. The incidents show that reliability of agentic file and database operations remains a live concern across competing AI tools.
Anyone who gives an AI agent write or delete access to personal or work files faces the same risk of unintended loss seen here. Keeping backups in a synced service such as Dropbox, as in this case, is what made recovery possible.
The takeaway
Agentic tools that edit files can fail in ways their users do not expect, so keep versioned backups before granting write access. Watch for any formal guidance from Anthropic on file-operation safeguards for agent deployments.
Further reading
For more on the reliability questions raised by agent tools, see the Artificial Intelligence section.
Source note: This article includes information reported by Tom's Hardware.
Live Poll
Would you trust an AI tool to edit your important work files?
