Syntony / Benchmarks

Governance
benchmarks.

Public datasets on organizational responses to security incidents and AI evaluations, beginning with PolBench.

Benchmark collection01 / Public evidence pilot
01 / GOVERNANCE RESPONSEv0.4 · October 2026

PolBench

Governance changes after incidents

PolBench covers 14 response episodes. Each record includes an evidence completeness score, the reported policy change and supporting sources. Sources were searched through October 20, 2026.

Open the interactive benchmark
PolBench
Finding
Decision
Ownership
Implementation
Validation
Five categories used to classify each response.
01

Source records

Every PolBench episode links to public records. Open a case to see which claim each source supports.

02

Missing information

Some actions may have no public record. Those checkpoints are labelled Not established.

03

Scope of comparison

This pilot compares specific response episodes. Its evidence coverage measure is not an organization-wide safety rating.

Corrections
and contributions

PolBench is a pilot dataset. Its methodology has not completed independent scientific validation. Send a public source link to suggest a correction or an additional case.

Contribute evidence