TENET AI Dev

The Firewall for AI Applications

Open-source defensive security middleware for LLM-powered apps.
Detect prompt injection, jailbreaks, and data extraction — in real time, with under 10ms overhead.

🌐 Website · 📦 Main Repo · 📖 Docs · 🤝 Contributing

TENET AI is a security plugin layer that sits between your application and any LLM API — OpenAI, Anthropic, Cohere, local models — intercepting every prompt before it reaches the model.

Your App  →  [ TENET AI ]  →  LLM API
                  ↓
            SOC Dashboard

Think of it as a firewall + intrusion detection system built specifically for AI.

As LLM-powered apps proliferate, they introduce attack vectors that traditional security tools don't cover:

Model-level guardrails can be bypassed. There is no unified security layer across providers. Security teams have no visibility. TENET AI fixes all of this.

TENET AI runs a four-stage pipeline on every prompt:

1. Intercept — Middleware captures outbound prompts before any LLM call
2. Analyze — Heuristic rules, ML classifier, and behavioral engine run in parallel
3. Decide — Policy engine issues a verdict: BLOCK / SANITIZE / FLAG / ALLOW
4. Learn — Analyst feedback and threat intelligence continuously improve detection

Total overhead: < 10ms.

More repositories coming as the project grows — SDKs, integrations, and deployment templates.

git clone https://github.com/TENET-DEV-AI/TENET-AI
cd TENET-AI
pip install -r requirements.txt
cp .env.template .env
docker-compose up -d

Integrate in 3 lines:

import tenet_ai

tenet = tenet_ai.Client(api_key="your-key")
result = tenet.check(prompt=user_input, user_id="u-123")

if result.blocked:
    return "⛔ Request blocked"

# Safe — call any LLM normally
response = openai.chat(user_input)

• ✅ Phase 1 (Now) — Ingest service, heuristic + ML detection, SOC dashboard MVP
• 🚧 Phase 2 — BERT-based models, behavioral analysis, multi-model support
• 🔮 Phase 3 — Multi-tenancy, RBAC, SIEM integrations (Splunk, Sentinel)
• 🚀 Phase 4 — Agent framework plugins (LangChain, AutoGPT), autonomous response

We welcome contributions of all kinds — detection model improvements, new attack datasets, integrations, documentation, and dashboard features.

See CONTRIBUTING.md to get started.

Found a vulnerability? Please read our Security Policy before disclosing.