NYC Local Law 144 · AI hiring tools
Bias audits your board, your auditor and your CISO can sign.
Fairspire™ calculates the selection rates, scoring rates and impact ratios that NYC Local Law 144 requires for automated hiring and promotion tools, by sex, race/ethnicity and every intersectional category, and drafts the summary you have to publish. Then it looks past the minimum: who your tool wrongly screens out, and why.
FairspireIf a tool helps decide who moves forward, New York City wants the numbers.
Local Law 144 covers employers and employment agencies that use an automated employment decision tool to substantially assist hiring or promotion decisions about candidates and employees in New York City. The Department of Consumer and Worker Protection has enforced it since July 5, 2023.
An independent bias audit
Done no more than one year before the tool is used, and every year after that.
A public summary
Rates, impact ratios, unknown counts, data source and audit date, posted on the employment section of your website.
Notice to candidates
At least 10 business days before use, including how to ask for an alternative process or an accommodation.
Penalties that add up
Up to $500 for a first violation and $500 to $1,500 for each one after it. Each day a tool is used without a valid audit counts on its own.
The required numbers, and the ones that should worry you.
Load an export of the decisions or scores your tool produced. Fairspire follows the DCWP rules step by step and shows its work, so your auditor can reproduce every figure.
The bias audit figures
Calculated for tools that decide and for tools that score.
- Selection rates for tools that advance or reject, and scoring rates for tools that score: the share of each category above the median score of the whole sample
- Impact ratios against the most selected or highest-scoring category, for sex, race/ethnicity (EEO-1 categories) and every intersectional category
- Unknown categories counted and reported, never guessed or compared
- Categories under 2% of the data can be left out of the ratios, with their counts still listed and a justification required
- A draft public summary as a web page and CSV, with a checklist of anything still missing before you publish
Why the numbers look the way they do
A clean sex table can hide a problem one level down.
- Wrongful rejections: qualified candidates the tool screened out, against recruiter review or your own written requirements
- Error parity: whether one group is wrongly rejected more often than everyone else, and whether the gap is statistically significant
- Identical profiles that received different outcomes
- Rejections without a specific reason, before a candidate or a regulator asks
- Findings mapped to Title VII, Illinois, New Jersey and the other rules for the states you hire in
In the example above, every sex and race/ethnicity ratio is above 0.80. Only the intersectional table shows Hispanic or Latina women at 0.70. A summary that stopped at the first two tables would look fine.
One audit, read five different ways.
Hiring tools touch the whole company. Fairspire gives each team the part it answers for.
A clear answer
Where the tool stands, what is being fixed and by whom, in a report you can take into the boardroom. The public summary carries your company's name, so you see it first.
No new platform
Works from a standard export of your applicant tracking system. Run it before the annual audit and again whenever the tool or the hiring volume changes.
The people behind the ratios
Which groups the tool screens out, which rejections were wrong, and which reasons are too vague to send. The facts to fix the process, not just to report it.
Nothing leaves
On-premise: desktop app or one container on your infrastructure. No vendor cloud, no telemetry. Exports and records hold aggregates only, never candidate rows.
Every figure traceable
Each step follows the DCWP definitions. Inputs are fingerprinted, results go to a tamper-evident ledger, and the method is documented for the independent auditor.
The law wants an independent auditor. We make their work faster.
Fairspire does not replace the independent bias audit, and it never claims to. It works on both sides of it.
Know before the auditor does
- Check a tool before you buy it, launch it or renew the audit
- Monitor it between annual audits, so a drift does not wait a year
- Hand your auditor clean, documented numbers and the data checklist
The calculations, done the same way every time
- Run the audit inside your client's environment: their data never reaches you or us
- Selection and scoring modes, intersectional tables and unknown counts out of the box
- Draft summaries you review, edit and sign as the auditor
Before you ask us.
Can Fairspire perform our Local Law 144 bias audit?
The law requires the bias audit to be conducted by an independent auditor. Independent auditors can use Fairspire to run and document the calculations, and employers use it to check their tools before the audit and to monitor them between audits. Every output says so.
What data does it need?
One row per candidate or employee the tool assessed: the tool's decision or score, and sex and race/ethnicity as your organization collected them, for example through voluntary self-identification. A standard CSV or JSON export from your applicant tracking system works. Missing or undisclosed values are counted as unknown, as the rules require.
Our hiring tool gives scores, not yes or no. Does that work?
Yes. For tools that score, Fairspire calculates the scoring rate the rules define: the share of each category that scores above the median score of the whole sample, and the impact ratio against the highest-scoring category.
Does candidate data leave our network?
No. Fairspire runs on your computers or your servers, with no vendor cloud and no telemetry. The summary it drafts and the records it keeps contain aggregate numbers only, never candidate rows or IDs. See how we protect your data.
Is an impact ratio below 0.80 a violation?
Local Law 144 requires the rates and impact ratios to be calculated and published; it does not set a pass mark. The 0.80 four-fifths benchmark comes from the EEOC Uniform Guidelines and is shown as context. A low ratio is a signal to investigate, and other laws such as Title VII may still apply. Try the math with our free four-fifths rule calculator.
We don't hire in New York City. Is this still useful?
Yes. The same engine audits hiring decisions against Title VII and the state rules that apply where you hire, such as Illinois and New Jersey, and it works for decisions made by people as well as by software.
Start with one hiring tool and your own data.
A paid pilot runs the Local Law 144 calculations and the full decision audit on one of your tools, on your infrastructure, with us guiding your team remotely. You keep the report.
- Sales sales@blackodix.com
- General hello@blackodix.com
- Security security@blackodix.com