model-armor

A chatbot for testing Google Model Armor LLM safety & security tool.

Autonomous Security & Zero-Trust Governance Guardian for Enterprise AI Agent Fleets powered by Gemini & Google Cloud. Built for #AllThingsAgenticHackathon.

🛡️ OpenClaw hook for Google Cloud Model Armor — LLM input/output sanitization (prompt injection, PII, harmful content)

Bastion — a governed institutional-agent fleet for event-driven access review. Technical report (Nov 2026): https://doi.org/10.5281/zenodo.24120596. Fleet retired 9 Nov 2026.

Event-aware contract intelligence for catastrophe reinsurance: deterministic payouts, verified clause citations, and an AI Courtroom that prices disputed exclusions.

Centralized AI security & governance platform integrating Google Cloud Model Armor, Sensitive Data Protection (DLP), and contextual defense

Sovereign, cloud-agnostic AI Governance reference architecture: Capability Advisory Matrix, automated PyRIT Red Teaming, and multi-cloud Model Armor runtime guardrails across AWS, GCP, and Azure.

Tutorial and proof of concept for Antigravity Lifecycle Hooks integrated with Google Cloud Model Armor runtime guardrails

AI safety proxy for Vertex AI. Screens prompts and responses for sensitive data and policy violations.

Terraform guardrails for generative AI on Google Cloud: model allowlist, predict-only identity, Model Armor, VPC-SC, audit and tamper alerts. 29 controls mapped to OWASP LLM, NIST AI RMF and 800-53.

Independent benchmark of guardrails for a tool-using AI agent on Google Cloud: Model Armor, rules and LLM guards on 999 synthetic scenarios

Attacking and securing a Vertex AI RAG app on Google Cloud, one control at a time. No controls: 17 of 20 attacks leaked PII. Model Armor plus DLP: zero leaks, with harmless questions still answered.

The agent was compromised. The payment was not.

A governed fleet of AI agents that runs municipal permit casework end to end on Google Cloud, and cannot issue a permit without a named human.

Apigee X AI Gateway with 3-Layer Defense-in-Depth Guardrails (Static Guardrails, Google Cloud Model Armor, and NVIDIA NeMo Guardrails) in front of Vertex AI Gemini 3.5 Flash Lite with real-time SSE streaming.

A governed fleet of fund-accounting agents that clears NAV reconciliation exceptions inside the valuation window. Gemini 3.7 Flash + ADK on Google Cloud. All Things Agentic Hackathon, Track C.

🛡️ PolicyPilot: guardrailed agentic RAG for HR policy on Google Cloud. Vertex AI Gemini, Qdrant hybrid search, Jina rerank, Model Armor, semantic cache, LiteLLM failover, LangSmith evals, Cloud Run.

Reference Google ADK Python agent template decoupling Model Armor binary prompt blocking from Cloud SDP in-place PII redaction for Gemini Enterprise.

AI governance console for Gemini Enterprise, Agent Platform, Model Armor, NotebookLM and Antigravity on Google Cloud. One-command Cloud Run deployment.