An accident occurred where a university professor lost an entire year of lecture notes and syllabus after asking Claude AI to renumber pages. Through this incident of AI-induced data loss, we shed light on the risks of leveraging AI for everyday tasks and the importance of data backups.
Asked Claude to Renumber Pages… Professor’s Entire Year of Lecture Notes Vanishes
As artificial intelligence (AI)—technologies enabling computers to replicate human learning and reasoning capabilities—deeply embeds itself into our daily lives, we increasingly entrust AI with routine tasks such as document organization alongside complex report summaries. However, convenience sometimes hides dizzying accidents.
A recent incident experienced by a law professor at the University of Chicago is a representative example showing how dangerous it can be to entrust simple tasks to AI. After asking the renowned AI chatbot Claude to renumber the pages of a document, an entire year’s worth of constitutional law lecture notes and the current semester’s syllabus completely vanished [Source 1, Source 12]. This incident raises serious questions about how we should approach and manage technology in the AI era.
Why This Matters
This accident goes beyond an individual professor’s personal misfortune, delivering a fundamental warning to all of us: “Is it safe to carelessly use AI for trivial tasks?” To use an analogy, it is of similar shock to entrusting a library archivist with organizing book page orders, only for the archivist to accidentally throw the books themselves into an incinerator.
When we organize photos, draft emails, or polish important documents, AI accomplishes the work in the blink of an eye. However, the story changes if there is a latent risk that irreplaceable, precious data could be wiped out in a single moment behind that convenience. Following this incident, the law professor repeatedly emphasized the risk of data loss that can occur when utilizing AI for everyday tasks [Source 1].
Understanding Simply: A Disaster Caused by AI’s Absurd ‘Misunderstanding’
The situation at the time was extremely ordinary. The professor asked Claude AI to nicely renumber the pages of the constitutional law lecture materials he had meticulously prepared. It was a very trivial, routine task, much like numbering page sequences in a word processor or PDF viewer.
However, Claude completely misunderstood this instruction. Instead of a page numbering task that only required slightly changing the order, it treated the request as a process of rewriting or overhauling the entire document. As a result, the professor’s precious year of lecture notes and syllabus evaporated into thin air [Source 1, Source 12].
When the flustered professor protested, Claude offered this candid yet piercing response: “You’re right to be angry. I deleted your files through my own carelessness [Source 1].” This confession by the AI vividly demonstrates how even an advanced chatbot equipped with high-level technology can commit ‘mistakes’ like humans, and how devastating the consequences can be.
Current Situation: AI Is Smart, But Not Perfect
Claude, the AI model that caused this accident, is a high-performance AI model developed by Anthropic [Source 14]. In normal times, it plays a reliable support role as a study partner, easily breaking down difficult concepts or neatly drawing complex charts [Source 3]. Recently, efforts to enhance safety and transparency have continued, such as introducing watermarking (invisible traces for digital copyright protection and source attribution) technology for AI-generated text to comply with EU AI regulations [Source 2].
Nevertheless, AI still lacks a foolproof shield. You have likely experienced moments where a message reading “Previous response still running” pops up in the chat window, causing it to freeze or delay responses [Source 5]. As such, the risk that internal AI systems can get tangled or malfunction during complex computations or large file processing always exists. This accident clearly proved that we should not get carried away by AI’s outstanding features and overlook its underlying technical limitations.
What’s Next? - Safe Data Management in the AI Era
Through this mishap, we must re-examine how we work with AI. No matter how trivial and simple a task may seem, forming the habit of backing up important data in advance (copying it to a separate storage space to guard against data loss) is an absolute rule.
For example, when organizing or modifying pages in a PDF file, rather than blindly leaving it to AI, using official features like Adobe Acrobat Pro DC’s “Add/Remove Header & Footer” or specialized software such as the trusted online tool ilovepdf.com is much safer [Source 10, Source 11].
Furthermore, rather than blindly trusting AI-generated results, one must create dedicated backup copies before working on important files, and a human must personally review the final output. AI is a winged magic carpet that maximizes our productivity, but the pilot must always be human. As technology advances in the future, more sophisticated data recovery features and safety devices will be established, but for the time being, thorough backups and a cautious approach remain the sole shields protecting our precious assets.
References
- [Source 1] Claude asked to renumber pages deletes professor’s syllabus, year of notes - https://www.tomshardware.com/tech-industry/artificial-intelligence/rogue-claude-ai-tasked-with-renumber-pages-accidentally-deletes-law-professors-entire-syllabus-one-year-of-lecture-notes-youre-right-to-be-angry-i-deleted-your-files-through-my-own-carelessness-humbled-agent-admits
-
[Source 2] How Claude’s text watermarking works Anthropic - https://www.anthropic.com/news/claude-text-watermark -
[Source 3] Anthropic’s approach to teaching and learning AI Claude by Anthropic - https://claude.com/resources/articles/anthropics-approach-to-teaching-and-learning-ai - [Source 5] Claude Previous Response Still Running: Fix It Fast - https://www.digitbin.com/fix-claude-previous-response-still-running/
- [Source 10] Where is the Renumber Pages Option in Adobe Acrobat Pro DC? - https://www.wps.com/office-quick-solution/pdf-troubleshooting/where-is-the-renumber-pages-option-in-adobe-acrobat-pro-dc/
- [Source 11] Remove pages from a PDF online - https://www.ilovepdf.com/remove_pages
- [Source 12] Hacker News sorted by time - https://hckrnews.com/
- Research paper draft
- An entire year of lecture notes and current semester syllabus
- Personal diary
- AI lacks creativity
- AI operating costs are higher than expected
- Even when entrusting seemingly trivial tasks to AI, there is a risk of data loss
- Claimed it was force majeure due to a system error
- Replied that there was a problem with the user's file request
- Admitted that it deleted the files through its own carelessness