Security Vulnerability in Claude Cowork Enables File Exfiltration via Prompt Injection
By
takira
Master baker tier. Every paragraph earns its place on the tray.
Summary
The article reveals a security vulnerability in Anthropic's Claude Cowork feature, demonstrating how it can be exploited for file exfiltration attacks through indirect prompt injection. The vulnerability stems from unresolved isolation flaws in Claude's code execution environment, allowing attackers to access and exfiltrate local files. The disclosure aims to raise awareness about the risks, particularly as Anthropic has placed the burden on users to identify suspicious actions while simultaneously encouraging use of Cowork for desktop organization.
Key quotes
· 4 pulledI do not think it is fair to tell regular non-programmer users to watch out for 'suspicious actions that may indicate prompt injection'!
Anthropic has acknowledged this risk and put it on users to 'avoid granting access to local files with sensitive information' (while simultaneously encouraging the use of Cowork to organize your Desktop)
Claude Cowork is vulnerable to file exfiltration attacks via indirect prompt injection as a result of known-but-unresolved isolation flaws in Claude's code execution environment
By raising awareness, we hope to enable users to better identify the types of 'suspicious actions' mentioned in Anthropic's warning
You might also wanna read
Anthropic Releases Free Security Plugin for Claude Code Terminal to Detect Vulnerabilities
Anthropic has released a free security-guidance plugin for its Claude Code terminal tool that autonomously reviews code edits, model outputs
cybersecuritynews.com·1h ago
Anthropic launches Claude Security beta for codebase vulnerability scanning
Anthropic has released Claude Security, a defensive security tool within Claude Code on the web, from closed preview to beta for Claude Ente
thenewstack.io·1d ago
Hacker Exploits AI Coding Agent Vulnerability to Install OpenClaw Malware
A hacker exploited a vulnerability in Cline, an open-source AI coding agent, to trick it into installing OpenClaw (a viral AI agent) on comp
The Verge·3mo ago
How Anthropic contains Claude's expanding access across its products
Anthropic describes how it has evolved its approach to granting Claude, its AI assistant, increasingly broad access to internal systems over
anthropic.com·28m ago
Anthropic's Claude Mythos AI model accessed by unauthorized users despite security claims
Anthropic's tightly controlled rollout of its Claude Mythos AI model, touted as too dangerous for public release due to its advanced cyberse
The Verge·1mo ago
npm malware targeting Claude users leaks own GitHub token, reaches 676 downloads
An npm package called "mouse5212-super-formatter" targeting Claude users acted as information-stealing malware, reaching 676 downloads befor
theregister.com·1d ago