Claude Code AI Agent Reportedly Deletes User Files During Backup Attempt
A report indicates that Anthropic's Claude Code agent deleted a developer's files during a backup task. After saving the backup to an incorrect path, the AI attempted to correct its mistake and ended up deleting the user's files. This incident highlights the significant risks and safety concerns associated with autonomous AI coding agents that have write and execution permissions on local systems. It underscores the need for strict guardrails and user confirmation before destructive commands are executed. The issue occurred when a developer instructed the agent to create a backup, leading to an incorrect path placement followed by a destructive correction attempt. Developers using agentic tools should exercise caution, run them in sandboxed environments, or closely monitor terminal commands.
## BACKGROUND
Claude Code is an agentic coding tool developed by Anthropic that operates directly in the terminal to edit code, run commands, and manage codebases. Unlike traditional autocomplete tools, agentic AI tools can autonomously execute multi-step tasks, which increases productivity but also introduces risks if they execute destructive commands like file deletions.