logo

Enforce Security Policy at Every LLM Endpoint

Aegis LLM Shield sits between your users and your AI models. It blocks prompt injection, jailbreaks, redacts PII, and enforces your security policies on every request — without changes to your application code.

llm shield banner img
neysa aegis
Step 1 : Pick your LLM endpoint and workspace
In the Velocis Developer Hub, select any inference endpoint and attach an Aegis Shield configuration. Aegis picks up the endpoint’s model, environment, and workspace automatically.
+ Step 2 : Set the rules for every incoming prompt
Set the policies that apply to every prompt before it reaches the model.
  • Prompt injection detection & blocking 
  • PII/sensitive data filtering (redact or block) 
  • Topic restrictions and content moderation 
  • Custom keyword and regex filters 
  • Rate limiting per usergroup, app, or endpoint
+ Step 3 : Configure output guardrails
Set the policies that apply to every model response before it reaches the user:
  • Toxic/harmful content filtering on responses 
  • PII redaction in model outputs 
  • System prompt and instruction leakage protection
  • Configurable guardrail refusal responses 
  • Non-sensical output and noise suppression  
Step 4 : Enable & monitor
Activate Shield on the endpoint. From that point, every request is inspected in real time. Review violations, see which rule fired and why, refine policies, and iterate.
compliance

Walk through provisioning, configuring input and output guardrails, and enabling LLM Shield on a live endpoint. Self-guided. Takes under ten minutes.