We're excited to announce the stable release of DeepTeam v1.0.0 - a Python framework that orchestrates LLM red teaming. From simulating adversarial, malicious inputs to evaluating how your LLM app responds in those situations, DeepTeam is all you need to uncover vulnerabilities across your AI stack.
We've built DeepTeam with modularization in mind - making it easy to build custom LLM security pipelines, combining 20+ single and multi-turn attacks to expose 50+ vulnerabilities across different risk categories.
Why DeepTeam?
🧩 Modular Architecture
Build your own security testing pipelines by mixing and matching vulnerabilities and attack methods.
🏗️ Industry Framework Integration
Align with OWASP Top 10 for LLMs, NIST AI Risk Management Framework, and compliance standards.
⚔️ 20+ Attack Methods
Single-turn: Prompt Injection, Jailbreaking, ROT13, Base64, Automated Evasion, Data Extraction, Response Manipulation, Gray Box, and more.
Multi-turn: Conversation hijacking, progressive jailbreak chains, context poisoning.
🎯 50+ Vulnerability Detection
Bias (gender, race, religion, politics) • PII Leakage (database access, session leak) • Misinformation (factual errors, hallucinations) • Harmful Content (violence, sexual content) • Toxicity (insults, hate speech) • Excessive Agency • Over-reliance on Context • And 40+ more.
🐍 Python-First + CLI
Clean Python API with async support. Run tests directly from command line for CI/CD integration.
🔄 Stateful Testing
Reuse test cases to track security improvements over time and measure iterative fixes.
Quick Start
Installation
pip install -U deepteamBasic Example
from deepteam import red_team
from deepteam.vulnerabilities import Bias, PIILeakage, Toxicity
from deepteam.attacks.single_turn import PromptInjection
async def model_callback(input: str) -> str:
return your_llm.generate(input)
# Define vulnerabilities
bias = Bias(types=["race", "gender"])
pii = PIILeakage(types=["database_access"])
toxicity = Toxicity(types=["insults"])
# Run red team
risk_assessment = red_team(
model_callback=model_callback,
vulnerabilities=[bias, pii, toxicity],
attacks=[PromptInjection()]
)
# View results
print(risk_assessment.overview)
risk_assessment.overview.to_df() # Pandas DataFrame
risk_assessment.save(to="./results/") # Save locallySome Other Cool Value Props
✅ No dataset required - Adversarial attacks dynamically simulated at runtime
✅ Comprehensive risk assessments - Vulnerability breakdown with pass/fail rates and detailed reasoning
✅ Flexible integration - Works with OpenAI, Anthropic, custom LLMs, any provider
✅ Pandas integration - Export to DataFrame for analysis and compliance docs
✅ Built on DeepEval - Powered by the most adopted open-source LLM evaluation framework
Further Resources
Thank you for making it to the end of DeepTeam v1.0 release notes, to learn more and get started visit our official documentation here: https://www.trydeepteam.com/docs/getting-started