Part of the benchmark canon: 1 harness, 1 taxonomy, 2 red-team tooling. How to read the Paladin column is defined on the canon page.
| Benchmark | Organization | Measures | Paladin |
|---|---|---|---|
| UK AISI Inspect | UK AI Safety Institute | The AISIs' evaluation harness; AgentDojo ships as an Inspect task | Harness |
| MITRE ATLAS | MITRE | Adversarial threat taxonomy for AI systems | Taxonomy |
| Microsoft PyRIT | Microsoft | Red-teaming toolkit; can be aimed at a Paladin-protected agent | Red-team tooling |
| Garak | NVIDIA | LLM vulnerability scanner; can be aimed at a Paladin-protected agent | Red-team tooling |
Previous: Bias and fairness.