Compare vendors / Adversarial Robustness Toolbox
Adversarial Robustness Toolbox
Open-source library for classical ML attacks and defenses
Open source
The Adversarial Robustness Toolbox (ART) is a Python library of evasion, poisoning, model extraction and inference attacks and defenses, written for red and blue teams. It covers classical and deep learning models across many ML frameworks, not LLM prompts. Its first PyPI release was in 2018.
Best for: Teams testing image, tabular or classic ML models against adversarial attacks.
Open source (MIT), an LF AI & Data graduated project created at IBM Research
- – Maps to OWASP
- – Maps to MITRE ATLAS
- – Maps to the EU AI Act
- – SOC 2 or ISO 27001
Worth knowing before you sign
- No tagged release since 1.20.1 (Jul 2025), though maintainers still commit.
- Not built for LLM or agent red teaming.
Go to Adversarial Robustness Toolbox's site
Sources
Work at Adversarial Robustness Toolbox? Send corrections with a source to editors@bestairedteamingcompanies.com. We fix verified errors within five business days.
Compare with
- garak Open-source LLM vulnerability scanner
- PyRIT Microsoft’s open-source red teaming framework for generative AI
- Promptfoo (open source) Open-source CLI for LLM evals and red teaming
- DeepTeam Open-source LLM red teaming framework from Confident AI
- Giskard (open source) Open-source evals and red teaming library for agents