AI Security Gateway careers

AI Security Gateway is the control layer between your application and any LLM provider. Every prompt is scanned for PII, secrets, and prompt injection attacks before it reaches the model. If the safety layer goes down, requests are blocked, never forwarded unscanned.

Every LLM application we audited had sensitive data flowing directly from user prompts to third-party AI providers, unfiltered. Emails, SSNs, credit cards, API keys — all in plaintext. Most teams don't discover this until a compliance review.

What we built:
• PII redaction across 30+ entity types (SSNs, emails, credit cards, API
keys, GitHub tokens, AWS credentials, street addresses, crypto wallets,
and more) — detected and redacted in under 50ms
• Prompt injection blocking — jailbreaks, DAN variants, SYSTEM OVERRIDE
impersonation, ignore-previous attacks, encoding exploits
• Per-project spend enforcement with hard 402 stops before overspend
• Smart cost routing across 300+ models from 8+ providers (OpenAI,
Anthropic, Google, Groq, Mistral, DeepInfra, Together, xAI)
• OCR-powered image scanning — catches PII in screenshots and documents
before vision models see them
• Metadata audit logs — every request logged with provider, latency, PII
flag, cost, and token counts without storing prompt content
• Semantic caching — For identical DLP-cleaned prompts.
• Webhook alerts — HMAC-signed push notifications for PII blocks, injection attempts, budget exhaustion, and loop detection. Slack, PagerDuty, or any SIEM
• EU AI Act Article 12 audit trails — hash-chained, tamper-evident records.


OpenAI SDK compatible. Change your base_url. Works with LangChain,
LlamaIndex, and any framework that accepts a base URL.

3 ways to use it:
• Self-host via Docker (Apache 2.0, free forever)
• Managed cloud (1M free credits, no credit card required)
• Hybrid VPC — compiled Go proxy in your network. Prompts never leave your VPC

Designed for teams building GDPR, HIPAA, and SOC 2-compliant AI applications.