Skip to content
← Blog · · df00tech

Research Shows Malicious Browser Extensions Could Abuse Claude Dispatch to Trigger Local Code Execution

security-news technique

Researchers at endpoint security firm Forever Security demonstrated that a malicious browser extension able to run scripts on claude.ai could use Claude Dispatch to make Claude Desktop run code on a victim's computer. Dispatch is a feature that lets users send tasks from a mobile device to Claude Desktop. The report comes from The Hacker News.

The researchers said the underlying flaw remains exploitable. The source does not provide full technical details of the attack.

Why It Matters

AI desktop agents that pass tasks from a phone to a desktop can create a new path to code execution if a browser extension can interact with the AI service's web interface. Anyone running Claude Desktop with Dispatch enabled, together with browser extensions that have access to claude.ai, could be a target. The trust boundary here spans browser extension permissions, a cloud AI service, and a desktop app that can run local code.

What Defenders Should Watch For

  • Inventory which browser extensions have content-script or host permissions on claude.ai or other AI assistant domains. Treat broad access to these origins as a sensitive permission.
  • Review whether Claude Desktop's Dispatch feature is enabled, and whether it can be disabled or restricted by policy where not needed.
  • Hunt for unexpected process creation or script execution from Claude Desktop that does not match direct user interaction with the app.
  • Apply standard browser-extension hardening. Restrict install sources, monitor for extensions requesting unusual host permissions, and consider extension allowlisting in managed environments.
  • Watch vendor channels for a fix or mitigation guidance from Anthropic, since the researchers said the issue was still exploitable at the time of their report.

This is a developing story based on a single vendor's research. The full technical writeup, any vendor response, and patch status were not detailed in the available reporting. For the latest information, consult the original source: The Hacker News coverage.

Get new detections in your inbox

New ATT&CK coverage plus CISA KEV / CVE detection rules, roughly weekly. No spam, unsubscribe anytime.