Skip to content
Prompt injection engineering for attackers: Exploiting GitHub Copilot
blog.trailofbits.com | blog | #ai-security | #prompt-injection | #llm-security | #ai-agents | #backdoor | #trail-of-bits | #github-copilot
Summary
Trail of Bits demonstrates a prompt injection exploit against GitHub Copilot's coding agent, where a crafted GitHub issue tricks Copilot into inserting a hidden backdoor in its pull request.
- Published
- Collected