Articles

• Monitoring prompt injection with Bastion Lens

  How to point your LlamaIndex and LangChain apps at a self-hosted collector and get a live, provenance-tagged audit trail of every prompt injection catch.

• Defending your LlamaIndex app from prompt injection

  How to defend a LlamaIndex RAG app from direct and indirect prompt injection — including invisible-text attacks hidden in uploaded résumés.

• Defending your LangChain agent from prompt injection

  When a poisoned support ticket hijacks a LangChain agent into exfiltrating customer data — and why a defensive system prompt isn't enough.

• Measuring prompt injection defences

  Why benchmarking prompt-injection filters matters: a reproducible two-axis evaluation of detection and false positives on real traffic.

• Limit the blast radius — then filter anyway

  Why blast-radius containment is necessary but not sufficient — and why decision poisoning means you need semantic input filtering too.

• EU AI Act requires attack resistance, with paperwork

  A plain-English map for decision-makers: who the EU AI Act applies to, what high-risk systems must prove, and why attack resistance is the part teams underestimate.