Daniel Miessler • 11/24/2025

Thoughts on Prompt Injection OPSEC

This article is a detailed rebuttal to a blog post arguing that prompt injection strings should be treated like zero-day exploits and kept secret. The author contends that AI red teaming is not 'security theater' and that sharing attack methodologies, even un-patchable ones like prompt injections, is crucial for developing mitigations and improving overall AI system security in a rapidly adopting landscape.

0 comments

#cybersecurity #prompt injection #ai security