Vanguard · LLM Proxy & AI Protection Gateway
Your employees use AI. Your secrets stay inside.
Vanguard is a real-time LLM proxy that intercepts outbound prompts before they reach external AI models like ChatGPT, Claude, or Gemini. It redacts API keys, email addresses, internal IP addresses, personal identifiers (SSN, NIR), and financial data in under 1 millisecond. After the model responds, Vanguard restores the redacted values in-flight so the user sees their original data. Zero configuration changes needed on the AI provider side.
What Vanguard protects
API Keys & Tokens
Detects and redacts API keys, bearer tokens, and service credentials embedded in prompts. Prevents accidental leakage to external AI models.
PII & Personal Data
Identifies email addresses, phone numbers, SSN/NIR, passport numbers, and other personal identifiers. Redacts them before the prompt leaves your network.
Financial Records
Detects bank account numbers, credit card numbers, IBAN, and financial amounts. Prevents sensitive financial data from being sent to external AI services.
Internal Infrastructure
Redacts internal IP addresses, hostnames, database connection strings, and architecture details that could expose your infrastructure if sent to external AI models.
Vanguard vs. DIY LLM Proxy
| Aspect | DIY Proxy | Vanguard |
|---|---|---|
| Latency | 50-200ms typical | Sub-millisecond |
| Pattern coverage | Regex-based, limited | ML-enhanced detection |
| In-flight restoration | Rarely implemented | Full restoration |
| Maintenance | 5x cost of commercial | Managed service |
| NIS 2 / RGPD | Not designed for | Built for compliance |
Part of Range Zero
Vanguard is the LLM privacy gateway of Range Zero. Combined with Atlas (attack surface) and Paladin (security drills), your organization gets end-to-end security coverage.
Explore Range Zero