0DIN

0DIN is an AI security testing platform backed by Mozilla that focuses on identifying and validating vulnerabilities in AI systems, including chatbots, AI applications, and agentic workflows. The platform operates through a researcher-driven model, drawing on a community of over 2,300 security researchers who submit real-world exploits targeting live AI models. Submitted attacks are verified by the 0DIN team, reproduced across multiple models, and compensated via bounties before being incorporated into the platform's probe library. The core product suite includes three main offerings: - Scanner: A turnkey AI security testing tool featuring a probe library, scheduled scans, dashboards, custom probe import, PDF reports, and SIEM export. Targeted at large enterprise and regulated organizations running structured red-team programs. - AI Vulnerability Intelligence: Curated, versioned Probe Packs and an intelligence feed delivered in JSONL/YAML formats compatible with tools like PyRIT and Garak. Designed for security teams running their own tooling who want a researcher-validated probe feed. - Defense: Embedded detection tooling for prompt-based attacks and agent threat hunting, including detection signatures and suspicious-interaction scoring for integration into existing platforms or products. 0DIN tests for threats such as prompt injection, system prompt extraction, context poisoning, tool-call abuse, business-logic bypass, and multi-turn social engineering. Findings are mapped to OWASP LLM Top 10 and MITRE ATLAS frameworks, and the platform generates audit-ready evidence packets formatted for SOC 2, ISO 27001, and NIST AI RMF. Target users include CISOs, red teams, AppSec engineers, GRC teams, and platform vendors embedding AI detection capabilities.