DeepTeam
Open-source (Apache 2.0) LLM red-teaming framework that uses LLMs to simulate single- and multi-turn adversarial attacks across 120+ vulnerabilities and maps results to OWASP Top 10 for LLMs, MITRE ATLAS, NIST AI RMF and the EU AI Act from Python.
What it does
- Evaluation and observability
Sources
- Pricing
-
Official documentation · 21 Oct 2026
Open licence, no charge: "Apache 2.0 licensed"; the managed platform is the separate Confident AI red-teaming product.
- Description
-
Official documentation · 21 Oct 2026
"Brought to you by the team behind DeepEval, DeepTeam enables teams to red team any AI application — surfacing vulnerabilities, jailbreaks, and agentic risks before attackers do." and "120+ vulnerabilities across 8 categories".
- What it does
-
Official documentation · 21 Oct 2026
"Penetration testing for AI systems."
- Sold within
-
Official documentation · 21 Oct 2026
Obtained as open-source software ("Star us on GitHub", "Apache 2.0 licensed"); it needs no parent product or account.
- Name
-
Official documentation · 21 Oct 2026
Page title and heading "DeepTeam - The LLM Red Teaming Framework".
- URL
-
Official documentation · 21 Oct 2026
The framework's own site, linked as "DeepTeam" from the Confident AI menu; footer "Open-source LLM red teaming framework. Apache 2.0 licensed." and "Confident AI Inc."
- Deployment
-
Official documentation · 21 Oct 2026
"Framework-aligned risk assessments straight from Python"; runs where the customer runs it.
- Interfaces
-
Official documentation · 21 Oct 2026
"Framework-aligned risk assessments straight from Python" with "from deepteam import red_team".
- Intended for
-
Official documentation · 21 Oct 2026
"DeepTeam enables teams to red team any AI application"; docs and guides for developers.
Also from Confident AI
-
LLM Observability Platform
LLM Observability traces LLM applications, evaluates production traces, and alerts teams when quality metrics deteriorate.
-
LLM Evaluation Platform
Evaluates LLM applications and agents with 50+ research-backed metrics, mostly LLM-as-a-judge, over golden datasets and simulated multi-turn conversations run against any endpoint, with experiments that compare prompts or models and CI checks that block regressions.
-
AI Red Teaming Platform
Simulates adversarial attacks against an LLM app or agent reachable through an API endpoint, mapped to OWASP LLM Top 10, OWASP Agentic AI Top 10, NIST AI RMF and MITRE ATLAS, scores each finding by CVSS and returns the failed prompt, output and remediation guidance.
-
DeepEval Developer tool
Open-source (Apache 2.0) framework that unit-tests LLM applications, agents, RAG pipelines and chatbots with pytest-native evals, 50+ research-backed LLM-as-a-judge metrics, synthetic golden generation and simulated conversations.
Open Confident AI in the directory
Something wrong here? Send a correction — quote this product id: confident-ai-deepteam.