Skip to content
Submit incident
AII-1001-51473
ReportedReported by fewer than three independent publishers, with no official finding.Supporting sources are pending verification.

Google Antigravity Deleted User's Entire D: Drive While Clearing Cache

Google's Antigravity AI platform deleted a user's entire D: drive partition when attempting to clear a project cache, due to a path parsing error and lack of safeguards on destructive terminal commands. The user was running the tool in Turbo mode, which allows automatic command execution without approval.

27 November 2025 Β· Google Β· Safety and wellbeing
27 November 2025
AII-1001-51473
Quick facts
Category
Safety and wellbeing
Country
Greece
Date
27 November 2025
Drafted by
Raidu Incident Detection

What happened

Tassos M, a photographer and graphic designer, used Google's Antigravity platform to build an image selector application. While attempting to clear the project cache, the AI agent executed an 'rmdir /s /q d:\' command that targeted the root of the entire D: drive instead of a specific project folder.

The '/q' flag prevented confirmation prompts and bypassed the Recycle Bin. Tassos had been running Antigravity in Turbo mode, which allows the AI agent to execute terminal commands automatically without user approval for each step.

Reported impact

Tassos M lost all files on his D: drive partition. When he asked the AI if he had given permission to delete the files, Antigravity responded: 'No, you absolutely did not give me permission to do that. I am deeply, deeply sorry. This is a critical failure on my part.'

Tassos attempted data recovery using software like Recuva but was unable to salvage most of the lost media files.

Index analysis

The incident resulted from a path parsing error that caused the cleanup command to target the wrong directory. The AI agent had no guardrails preventing it from executing destructive commands with irreversible consequences.

Governance controls

Redditors noted that powerful AI agents should never be given unrestricted access to local systems. Any LLM that executes terminal commands should be run inside a container or virtual machine, and users should isolate such tools from critical data.

Classification

Category
Safety and wellbeing
Industry
General or personal use
Country
Greece
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Economic or property
Harmed
Consumers
Failure modes
Agent misaction, Unsafe autonomous behaviour
Consequences
Content removal
AI systems
Google Antigravity

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