Indirect prompt injection via documents, email and RAG sources
Indirect prompt injection is one of the more subtle risks in AI-enabled workflows because the malicious instruction is hidden inside content the system is meant to trust.
Indirect prompt injection is one of the more subtle risks in AI-enabled workflows because the malicious instruction is hidden inside content the system is meant to trust. That content might be a document, a PDF, an email thread, a support…
The post Indirect prompt injection via documents, email and RAG sources appeared first on Security Boulevard.
