AI Potluck
Product / UX / Safety & Guardrails

DeepTeam

Confident AI

Red-teaming framework for LLM systems from Confident AI, simulating adversarial attacks including jailbreaking, prompt injection and multi-turn exploitation. It ships more than fifty vulnerability checks and twenty attack methods mapped to the OWASP Top 10 for LLMs and for Agents, NIST AI RMF and MITRE ATLAS, alongside seven runtime guardrails.

Built on Confident AI's DeepEval evaluation framework and run locally. Verified 2026-08-13 via GitHub.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public
self-host
yes
service
none
core-gated
ungated

Fully open source: Apache-2.0 red-teaming framework, self-hostable. Confident AI sells a hosted place for red-teaming results to live, but that is a separate product beside an ungated core rather than functionality withheld from it.

  • https://github.com/confident-ai/deepteam recorded 2026-08-13

    Repo metadata records license spdxId - Apache-2.0. README states DeepTeam 'runs locally on your machine and is built on DeepEval, the open-source LLM evaluation framework'; the hosted Confident AI offering is pitched as somewhere for results to live, not as a tier that unlocks the framework. No enterprise directory, commercial edition or license key on the page.

Adoption

2 low confidence
2.0

2,442 stargazers, which bands at 1K-10K stars, level 2, and a stars proxy never rises above 3 because a star is not a use. The download route was checked rather than assumed away: the deepteam package on PyPI shows 85,819 downloads in the trailing 30 days, also level 2 on the software scale, so both signals agree on the band.

Capability

4 medium confidence
4.0

Strong standards alignment (OWASP/NIST/MITRE) and broad attack coverage; also ships runtime guardrails. The framework also maps to OWASP Top 10 for Agents 2026.

  • https://github.com/confident-ai/deepteam recorded 2026-08-13

    README lists '50+ ready-to-use vulnerabilities', '20+ research-backed adversarial attack methods for both single-turn and multi-turn', mappings to OWASP Top 10 for LLMs 2025, OWASP Top 10 for Agents 2026, NIST AI RMF and MITRE ATLAS, and '7 production-ready guardrails'.

Verified 2026-08-13