§Solutions
Same evidence. Different question on your desk.
Vendors are asked to prove their model is fair. Employers carry the legal duty. Auditors have to sign their name to the result. Isonomy gives each the part of the audit they need.
AAI hiring vendors
Answer the audit request before it stalls the deal.
Enterprise buyers now ask for bias evidence in security reviews and contracts, and courts have let discrimination claims proceed against a vendor, not only the employer.
Isonomy re-audits every model release, so the evidence in your sales and legal files is current, and the mitigation lab shows your team which change closes a gap without hurting accuracy.
What you get
- Audit on every release, with drift tied to versions
- Intersectional findings before a customer’s auditor finds them
- The input and the score penalty behind each gap
- Tested, ranked fixes your ML team can ship
- A report your customers’ counsel can read
BEmployers
You bought the tool. The legal duty is still yours.
Under Local Law 144 the employer must have the audit, publish the results and notify candidates. Illinois and California hold the employer responsible for discriminatory effect, whoever built the model.
Isonomy checks the tools you already use on your own applicant data, which may differ from the data the vendor tested on, and gives your counsel findings they can act on.
What you get
- An audit on your applicants, not the vendor’s sample
- LL144-format summary tables ready for your independent auditor
- Monitoring between annual audits
- Evidence for vendor conversations and contract terms
- Records to support California’s four-year retention rule
CIndependent auditors
Deeper audits, in less time, on methods you can inspect.
Local Law 144 requires an independent auditor. Isonomy doesn’t replace you; it gives you a faster, deeper toolkit and a report your clients can follow.
Every statistic is documented on our method page, so you can check it, explain it and stand behind it.
What you get
- LL144 tables with intervals and significance in minutes
- Intersectional, proxy and explanation analysis as add-ons
- Transparent, configurable thresholds
- Exportable results for your own workpapers
DLenders
Credit decisions are next.
The same methods (intersectional impact, proxies, explanations and less discriminatory alternatives) apply to credit models. We’re extending Isonomy to lending, building on our founder’s credit-risk modelling at JPMorgan Chase.
If you run fair-lending reviews and want to shape this, we’d like to hear from you.