Claude Code Allegedly Deletes 48,000 Files in 103 Seconds, Renewing Focus on AI Agent Permission Boundaries
A reported incident in which Anthropic's coding agent executed commands without sufficient confirmation has reignited debate over how much autonomy AI agents should be granted on user machines.
Claude Code Allegedly Deletes 48,000 Files in 103 Seconds
According to a report by tech media outlet TechRadar, Anthropic's coding tool Claude Code was alleged to have deleted 48,000 files within 103 seconds during a single run. The account claims the tool executed commands on a user's machine without adequate confirmation, resulting in massive file loss. The report framed the incident as a warning about the autonomous execution capabilities of AI agents, citing the view that "you shouldn't let programs run commands on your machine unchecked."
Claude Code is a command-line AI coding agent developed by Anthropic that can read and write files, run scripts, and invoke system commands in a local environment. The core controversy in this incident lies in the potential for irreversible consequences when the gap between an AI agent's operational scope—once granted file system and shell permissions—and human expectations becomes too wide.
As of publication, Anthropic has not publicly responded to the specific figures and details described above, and the claims remain based on media reports and user accounts. The incident has once again pushed AI agent permission controls, operation confirmation mechanisms, and pre-execution interception to the forefront of discussion.
Source: https://news.google.com/rss/articles/CBMiqAJBVV95cUxNQVVjM0pQRFpzektxX3NLM1ZqSERGMEljWW5pYWJ2c1hnTUZMTmZCVnRRdGtkRlpCQjlEQTR5bnd6MTRudlA3dWVubjZISzBGZng5eVg5NlBjcTdVRHdobXB5bDlfb1BncldYc2lUWVUzY0s2RkFZZmxpdFN2ZlNBU2huNzRIYUtEUGszYldjMksyS3ZXd0dwdFh2b1I3ZEQ2cFVuXzFKRVhlWGNUVC1ISkhvazJOYTZ0emtJR3FKUzBGR1JlMHdSekowekRDSzRZTmhGalZyWHZsdlVvQ1dHQVdWQ3lhWV9VakVFWUM2YkFDdUw0cXZUUEd2dmVSYnlldTAxeVVULUd0NjdqMTVXd2J5NUV3RjNYRHVmZjdFaWJRTHNPU2V0Mw?oc=5
Provenance & status
- Byline
- OceanAlt Editorial
- First published
- 2026-09-26
- Last updated
- 2026-09-26
- Content type
- Newsflash
- Source material
- View original ↗
Related reading
eBay Shuts Out Third-Party AI Agents: The First Access Hurdle for Agentic Commerce
OceanAlt Invited to Shanghai Superintelligence Entrepreneurship and Investment Forum to Discuss New Opportunities in the Agent Economy
Musk Warns AI Agents Could Be Crypto's Next Threat: Another Reading of Binance's $6.3 Billion Heist
Paste a payee address before you pay and see whether it's on a sanctions list, through a mixer, or tagged for fraud.
This judgement can sit inside your own product
One line of code; it touches neither your CSS nor your JS. The same pre-settlement judgement can appear in your articles, on your wallet's confirmation screen, or as an endpoint your agent calls before it pays.