NYC Local Law 144 · EU AI Act · CO SB21-169
Bias audits that hold up under cross-examination.
Independent algorithmic audits delivered as a notarized report — methodology spelled out in plain English, statistics in the appendix, defensible under audit by counsel and regulators.
What an audit covers
- Scope & population. Defining the impacted class precisely enough that the result is reproducible.
- Disparate-impact analysis. Selection rates by protected category, with confidence intervals and effect sizes.
- Provenance. Training-data lineage and feature-engineering decisions, traced to source.
- Remediation roadmap. What to change and what to monitor; tied to your deployment cadence, not ours.
- Public-facing summary. The plain-English version your candidates and HR team can read.