Governance Bench: Operational AI Agent Governance Benchmark

HUMMBL Governance Bench is the first benchmark designed to test operational governance for AI agents, focusing on what agents do rather than what they say. It evaluates whether AI agents can correctly interact with, reason about, and respect governance primitives that control execution boundaries. The benchmark was created by hummbl-hf and was last updated on June 21, 2026.

Use Cases

• Benchmarking AI agent compliance with operational governance primitives based on the described test focus
• Evaluating agent reasoning about execution boundaries as mentioned in the description
• Testing agent interactions with governance controls referenced in the benchmark's purpose

Strengths

• First benchmark for operational AI agent governance, a distinct focus from content safety
• Explicitly tests interaction with governance primitives as described

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

hummbl-hf

Freshness

Last updated 2026-06-21 22:51:14