deepteam

DeepTeam is an open-source, local-first LLM red teaming and penetration testing framework developed by Confident AI. Built on DeepEval, it simulates 20+ research-backed adversarial attacks—such as jailbreaking, prompt injection, and multi-turn exploitation—to uncover 50+ vulnerabilities (including PII leakage, bias, and security bypasses) in AI agents, RAG pipelines, and chatbots, while providing production guardrails to secure AI applications.

• 50+ Built-in LLM Vulnerability Checks
• 20+ SOTA Adversarial Attack Simulations
• Local Execution & LLM-as-a-Judge Evaluation
• Production-Ready Safety Guardrails
• Seamless Confident AI Platform Integration