DeepTeam
Python LLM red teaming framework that runs tests on your machine, saves local reports and uses model APIs for attacks and scoring. Apache 2.0.
DeepTeam is a Python framework that runs on your machine to test chatbots, AI agents and retrieval-augmented generation applications for security failures. It's for developers and security teams who want adversarial tests they can repeat as an application changes. Its model-powered attacks and scoring use the selected provider; calls to OpenAI or other cloud services send data off your machine.
Built on DeepEval, the framework generates attacks against chosen vulnerabilities without requiring a prepared dataset. Tests cover prompt injection, leaked personal data, SQL injection and authorization failures. Multi-turn methods such as Crescendo probe how an application responds as an attacker gradually escalates a conversation.
Agent-specific checks examine excessive authority, hidden instructions in retrieved content, misuse of tools and compromised messages between agents. Model judges return pass/fail scores with explanations, and teams can add their own vulnerabilities and attacks.
Risk assessments map tests to OWASP Top 10, MITRE ATLAS, NIST AI RMF and EU AI Act controls. BeaverTails and Aegis provide safety benchmark datasets. Reports can stay local as JSON, and GitHub Actions or GitLab CI can run repeated assessments.
The Apache 2.0 framework also includes input and output guardrails for prompt injection, privacy leaks, harmful content and hallucinations. Confident AI is a separate hosted platform for tracking assessments, monitoring production vulnerabilities and sharing reports.
Logotyper
