Control safety case for an AML investigation agent, mapped to SR 26-2, with a ControlArena evaluation of deterministic enforcement vs. LLM monitoring for AI agents.
-
Updated
Nov 6, 2026 - Python
#
controlarena
Here are 2 public repositories matching this topic...
An Evidence-Sufficiency-Card scorer and CI deployment gate for agent eval traces (Inspect, ControlArena, JSON) — reports which evidence is present, partial, or missing before a green eval is trusted as release evidence.
python ci-cd ai-safety ai-agents evidence llm evals llm-evaluation agent-evaluation inspect-ai deployment-gate controlarena
-
Updated
Sep 1, 2026 - Python
Add this topic to your repo
To associate your repository with the controlarena topic, visit your repo's landing page and select "manage topics."