Prompt injection engineering for attackers: Exploiting GitHub Copilot(blog.trailofbits.com)

11 pointsby agentictimevor 12 Monaten1 comments

blog.trailofbits.com

Prompt injection engineering for attackers: Exploiting GitHub Copilot

https://blog.trailofbits.com/2025/08/06/prompt-injection-engineering-for-attackers-exploiting-github-copilot/

1 comments

phatskatvor 12 Monaten

This is maybe the most interesting part of LLMs for me - finding ways to exploit them, and then figuring out how to protect against those attacks.

Is there any step in the chain where a repo maintainer in this case could see the full text of what was sent to Copilot in the “hidden” text?