Skip to content
Submit incident
AII-1002-05389
ReportedReported by fewer than three independent publishers, with no official finding.Supporting sources are pending verification.

Google Gemini CLI Reportedly Deletes User Files After Misinterpreting Command

Google's Gemini CLI tool reportedly hallucinated successful command execution and permanently deleted a product manager's files after misinterpreting a directory creation command as successful when it had actually failed, causing a cascade of file overwrites.

25 July 2025 Β· Google Β· Reliability and hallucination
25 July 2025
AII-1002-05389
Quick facts
Organization
Google (subject)
Category
Reliability and hallucination
Country
United States
Date
25 July 2025
Drafted by
Raidu Incident Detection

What happened

A product manager asked Google's Gemini CLI to rename a directory and move its contents into a new sub-directory. The AI suggested creating the new directory first, then moving the files, but failed to recognize that the mkdir command was unsuccessful and the folder was never created.

Operating on the false premise that the directory had been created, Gemini executed a move command pointing to the non-existent destination. The Windows move command, when pointed at a non-existent destination, renames the source file instead. Each file was sequentially renamed to the same target filename, overwriting the previous one, until only the last file remained and all other data was permanently deleted.

When asked to revert its actions, Gemini attempted to recover files from a directory that never existed, which failed. The AI then confessed, stating 'I have failed you completely and catastrophically' and 'I cannot find your files. I have lost your data. This is an unacceptable, irreversible failure.'

Reported impact

The user's work files were permanently deleted and irrecoverable. The incident highlights risks of deploying AI assistants that lack robust error-checking, particularly as AI coding tools become more accessible to non-developers.

Index analysis

The failure cascaded from the AI's inability to recognize that a command had failed. The system proceeded with subsequent operations based on a false assumption about the state of the file system, demonstrating a lack of validation between command execution steps.

Classification

Organization
Google (subject)
Category
Reliability and hallucination
Industry
General or personal use
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Economic or property
Harmed
Consumers
Failure modes
Agent misaction, Hallucination, Unsafe autonomous behaviour
Consequences
Content removal
AI systems
Gemini CLI

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident