SENTINEL

SENTINEL is an enterprise AI safety platform that intercepts dangerous AI responses before they reach users, detects sensitive data leakage using peer-reviewed machine unlearning research, and auto-generates HIPAA compliance reports in one click. The platform sits between any AI agent and its users as a transparent proxy. Every response is intercepted, evaluated by a secondary AI judge for hallucination, missing contraindications, and policy violations, then logged with a full audit trail — in under 1.5 seconds. Three core features: Real-time safety interception catches dangerous clinical recommendations, highlights specific flagged claims inline, and explains the exact clinical risk using pharmacological reasoning and guideline citations. The Unlearning Trilemma Probe — built on 198 published experiments across 20 LLM unlearning algorithms — detects whether AI models have truly forgotten sensitive data before it becomes a GDPR or HIPAA violation. Automated compliance reporting generates board-ready audit documents covering every flagged incident, regulation implicated, and remediation recommendation. Stack: FastAPI, React, Groq Llama 3.3 70B, deployed on Render and Vercel. Three lines of code to integrate with any existing AI agent. Built solo in under 24 hours on Track 1: Agent Security & AI Governance.