---
id: 1331
title: "Prompt injection to RCE in AI agents"
source_domain: blog.trailofbits.com
source_url: "https://blog.trailofbits.com/2025/10/22/prompt-injection-to-rce-in-ai-agents/"
published_at: "2025-10-22 11:00"
tags: ["ai-security", "rce", "prompt-injection", "sandboxing", "llm-security", "ai-agents", "argument-injection"]
phying_url: "https://phying.com/item/1331"
lang: en
---

# Prompt injection to RCE in AI agents

## Summary

Argument injection into pre-approved 'safe' commands bypasses human approval gates in AI agents, yielding RCE on three popular platforms; sandboxing and argument separation can limit the impact.

## Intelligence Metadata

- **Sources:** blog.trailofbits.com
- **Published:** 2025-10-22 11:00
- **Original:** https://blog.trailofbits.com/2025/10/22/prompt-injection-to-rce-in-ai-agents/
- **Phying News:** https://phying.com/item/1331