AIMoat

▶ SUMMARY

Security scanner for public AI chatbot integrations. Built to test how AI chat products fail in the real world: prompt injection exposure, system prompt leakage, unsafe response rendering, endpoint abuse, weak browser storage, missing security headers, and weak evidence around model behavior.

▶ HIGHLIGHTS

• → 8 security check modules covering chatbot, browser, and deployment risks
• → 30+ probe corpus including Policy Puppetry, Deceptive Delight, FlipAttack, and Typoglycemia
• → Detects Intercom, Crisp, Drift, Zendesk, HubSpot, and LiveChat deployments
• → A-F severity grading with masked evidence on free tier and full disclosure on paid

▶ STACK