Skip to main content
Content Analyzers are specialized plugins that inspect content from external sources before it reaches your AI application. They protect against indirect prompt injection attacks and prevent sensitive data exposure.

​
Available Analyzers

Document Analyzer

Scan uploaded files (PDF, DOCX, images) for PII, jailbreaks, and sensitive content

URL Analyzer

Fetch and analyze content from URLs for jailbreaks and PII

​
Why Content Analyzers?

Modern AI applications often process content from multiple sources:
  • Uploaded documents: Users attach files to their requests
  • External URLs: Users reference links that the AI should analyze
  • Embedded content: Documents may contain links to other resources
Each of these represents a potential attack vector:
ThreatDescription
Indirect Prompt InjectionMalicious instructions hidden in documents or web pages
Data ExfiltrationSensitive data embedded in external content
PII ExposurePersonal information in uploaded files

​
How They Work

​
Common Configuration

Both analyzers share similar configuration patterns:
SettingDescription
modeenforce to block threats, observe to log only
pii.entitiesList of PII types to detect
firewall.enabledEnable jailbreak detection
firewall.thresholdJailbreak score threshold (0.0-1.0)
firewall.credentialsNeuralTrust API credentials for jailbreak detection

​
Use Cases

  1. RAG Applications: Ensure retrieved documents don’t contain malicious prompts
  2. Document Processing: Validate user uploads before AI processing
  3. Web Scraping: Analyze fetched web content for security threats
  4. Compliance: Detect and log sensitive data in all content sources