Prompt injection & LLM jailbreaks
Incidents where adversarial input — direct or indirect — caused a generative model to override its instructions. Cross-references active lawsuits, applicable laws, and red-team vendors.
Start here
Informational
What is prompt injection?
How direct and indirect prompt injection works, with annotated examples.
Comparison
Prompt injection vs jailbreak
Distinguishing the two attack classes — and why mitigations differ.
Commercial
How to prevent prompt injection
Engineering and vendor controls — what works, what doesn't.
Spokes
Logged incidents
Vendor spoke
Compliance & monitoring vendors for prompt injection
Reviewed vendors building tools to prevent this failure mode — maintained in our sister directory.