AI Security Gateway

Open-source AI firewall- redact PII, block prompt injection

Visit Website
April 21, 2026 Open-source AI firewall — redact PII and block prompt injection before it reaches the LLM

AI Security Gateway (AISG) started when I found customer emails showing up

in our own LLM provider's logs. Not a breach — our app was sending them

in prompts without realizing it.

I looked for a self-hosted solution that was genuinely secure and

drop-in compatible with existing OpenAI SDK code. Couldn't find one.

Built it instead.

AISG (https://aisecuritygateway.ai/) is an OpenAI-compatible proxy that sits between your app and any

LLM provider. Every request is scanned for PII, secrets, and prompt

injection attacks before it leaves your infrastructure. Change your

base URL and API key — nothing else changes in your code.

I built AISG as a hardened, stateless security gateway and intelligent router that sits between your code and any provider (Groq, Together.ai, DeepInfra, Mistral AI, Anthropic Claude, OpenAI, Google Gemini, xAI Grok, and open-weights models).

It doesn't just proxy — it sanitizes in real time:

  • 🛡️ Multi-modal DLP: 28+ PII entities redacted in <50ms, including visual OCR for screenshots/images.

  • 💸 Smart routing: Always picks lowest-latency + best-price option across providers.

  • 📊 Governance: Per-project budgets, circuit breakers for agent loops, audit logs. Fail-closed by design.

Built solo on AWS.

  • Fail-closed by default — if PII Scanner is unreachable, requests are blocked, never forwarded unscanned

  • Auth by default — API key authentication enabled out of the box

  • No telemetry — zero external calls, no analytics, no phone-home

  • Secret scrubbing — structured logs automatically mask API keys and tokens

  • Rate limiting — token bucket per API key (default 10 req/sec)

  • Stateless — AISG do not persist, store, or “train” our models on the text of prompts or the content of AI responses.

Business model:
💸 The open-core. The self-hosted version is free and open source (Apache 2.0). https://github.com/aisecuritygateway

💸 The managed Cloud: adds Smart Cost Routing across 8+ providers, dashboards, multi-project DLP policies, and team management — starting with 1M free credits, Pro at $29/month.
https://aisecuritygateway.ai/

Early stage, bootstrapped, building in public.

Comment

About

As a software architect, I watched teams block LLM usage entirely due to shadow AI risks — engineers pasting secrets, PII, or credentials into prompts. So I built AI Security Gateway (AISG) , an OpenAI-compatible proxy.