The ADS rules under the Fair Employment and Housing Act took effect on 1 October 2025. They bind employers and the agents who screen for them, and nothing is filed with anyone at all: they make bias testing the evidence that answers a decision once it is challenged.
CALIFORNIA FEHA · ADS TESTING · IN FORCE
Nothing to file up front. Everything to show when challenged.
The rules bind your agents and vendors as well as you - in force since October 2025, and nothing is filed until a decision is challenged.
Automated decision systems
Anti-bias testing
Agents and vendors
Records
Adverse impact
Our promise
“A scorecard is an opinion. Bias testing is evidence.”
Every finding is written against a clause of the instrument itself - defensible line by line, to anyone who asks. The fee is fixed at $6,500, and nothing is charged until you approve it.
- Each additional tool
- $1,500
- Re-audit, same scope
- $4,200
- Renewal, every twelve months
- $5,500 locked
This costs more than the estate ladder because it is published: iDharma keeps your summary live for twelve months, with notice and renewal checks.
Request this auditThe case file
California’s ADS rules, in three chapters
Because nothing is submitted, nothing forces it onto a date. Teams hold a vendor’s test on someone else’s applicants, no inventory, and no record. None of it exists until it is asked for, and a test of the applicants you screened cannot be built after the fact, once they are gone.
Our review produces that evidence before it is asked for. We inventory every system that takes part in an employment decision, test adverse impact on your own applicant data, write down the method as well as the numbers, and log a decision against each finding we raise.
What is an automated-decision system?
A tool that takes part in the decision.
It need not make the final call. If it ranks, scores, screens or filters people, it is taking part.
- California employers whose tools touch a hiring decision
- Agents performing employment functions on their behalf
- Vendors doing the screening, not just selling the tool
- Employers outside California hiring for California roles
- When an automated employment decision is challenged
- Not before - there is no filing, registration or certificate
- Which is why the record has to exist before anyone asks for it
- And why later testing evidences the test, not the decision
Whose duty is it
The tool is theirs. The decision is yours.
The employer making the decision
FEHA places the obligation on whoever is making the decision. Buying the tool does not move it, and nor does letting somebody else run it: the regulations reach agents performing the employment function on your behalf, which is exactly what a screening agency does when it screens for you.
The people who built the tool
Often testing their model carefully, and often publishing the results. The difficulty is not the quality of that work - it is whose applicants it was run on, which is a different question from whether it was done well, and it is the one question nobody in the room ever thinks to ask the vendor before signing.
When their testing covers you
Only to the extent your own applicants were in it. If you have been running the tool on your own candidates and your data was not in that testing, you do not have testing about your use of it - however good theirs is, however large the sample was, and however recently it happens to have been run.
What most teams assume
“Our vendor ran a bias audit, so we’re covered.”
What the rules make relevant
It describes someone else’s hiring, not yours.
It is the most common gap we write up.
- Who it is for
- In-house talent teams
- Staffing & RPO
- HR-tech vendors
- Legal & compliance
- ATS and assessment platforms
Why the record matters
02
A role in New York City brings Local Law 144 to the same tool - an independent annual audit, a published summary of results, and notice.
03
A role in Illinois brings HB 3773 - notice that AI is being used, and discrimination by AI treated as a civil-rights violation in its own right.
04
Where two of them reach one role, the answer is one piece of testing designed to the stricter of the two - not two designed to satisfy each.
The 60-second check
Three questions. Then you’ll know.
No email. No signup. A starting point, not a determination.
Opens in a panel. Close it any time - your answers are kept.
California FEHA · ADS
Your scope check
The four moments
No deadlines, and four ways to be late.
Nothing here is measured against a date in a rule. Each of these is measured against something you did - which is why none of them can be caught up afterwards.
-
Classify
Before first useEvery tool that touches an employment decision, named and classified - including the ones you ruled out of it.
-
Test
On your own dataAdverse impact measured on your own applicants, not a vendor's reference population. Method recorded.
-
Re-test
On model changeA retrained model is a different tool. The calendar alone lets one go a year without anybody looking at it.
-
Keep
Before the challengeThe record has to exist before anyone asks. Testing run afterwards is evidence of a test, not the decision.
The trap
Because nothing gets filed, nothing forces the work to happen - so the record is usually commissioned after a decision has been challenged. Testing run then is evidence of a test. It is not evidence of the decision, which was made on a tool nobody had looked at.
Relevance & coverage
What the rules make relevant, what we ship
12 things a challenge can test, and the artefact that answers each one.
- Whose obligation it is The employer - and any agent acting for one
- A written determination of who owes the duty for each tool, so a vendor's testing is never mistaken for yours.
- Reliance on vendor testing Only to the extent your applicants were in it
- An assessment of whether your vendor's testing describes your use of the tool - and what must be re-run if it does not.
- Which systems were in scope And which were not, with the reasoning
- A named inventory of every tool that touches an employment decision, with the exclusions and why each was made.
- Whose data it ran on Yours, ideally - and said plainly either way
- The population behind each test stated on its face, rather than left to be assumed by whoever reads it later.
- What was measured Method, not just the outcome
- The characteristics tested, the comparison used, and the treatment of categories too small to say anything about.
- When it was run Against which model version
- Every test dated to a model version and a data window, because a test with no date cannot be shown to be recent.
- What you found Including the awkward parts
- Findings recorded as found. A record containing only clean results reads as a record that was curated.
- What you did next The response is evidence too
- Per finding: remediated, accepted with a stated reason, or open with an owner and a date against it.
- Small-category handling Where the numbers are too thin to speak
- The rule you applied, applied consistently, and its limitations written into the report rather than around them.
- Agents and vendors The regulations reach them expressly
- The evidence to ask each vendor for, and the contract language that gets it at renewal instead of at litigation.
- Where the other two reach the same tool NYC Local Law 144, Illinois HB 3773
- One piece of testing designed to satisfy the strictest instrument that reaches the role, rather than three.
- Keeping the record recent Recency is one of the things made relevant
- A re-test trigger on model and data change as well as on the calendar, and a named owner for the record itself.
The engagement
Hiring tools, independently tested
From CV screening to assessment scoring.
-
Inventory
Which tools touch an employment decision, and which you ruled out.
-
Test
Adverse impact on your own applicant data, method recorded as you go.
-
Sign off and keep
You see the draft first. Then the record, dated to the model it describes.
Struck in your favour
Why employers choose iDharma for FEHA
Genuinely independent
We build and resell no hiring tools of any kind, and take no fee tied to what the testing finds.
We sell no certificate
There is no accreditation scheme here, so nobody can be certified against it. We say so.
One review, three laws
California, New York City and Illinois all scope into one piece of testing, not three of them.
The record, not a score
You get the method, the findings and the response - the parts a challenge actually tests.
Four marks, struck on every record.
Deliverables
What you get
Concrete artefacts, each with a name and a format - you know what lands before you buy.
The testing record
The deliverable in one document, because there is nothing to file: every system in scope and every one ruled out, adverse impact across protected characteristics on your own applicants, the small-category rule applied and stated, and a logged response against each finding - dated to the model tested.
ADS inventory
Every tool that touches an employment decision, with its stage, its owner, its vendor and the in-or-out call recorded against each entry.
Statistical results
The underlying tables in a workbook, so your own analysts can reproduce every figure in the record, cell by cell, without asking us for it.
Method note
The characteristics tested, the comparison used and the small-category rule - written at the time rather than explained afterwards.
Findings and response log
Each finding with its decision beside it: remediated, accepted with a reason, or open with an owner and a date. Ranked by consequence.
Vendor evidence pack
What to ask each vendor for, and the contract language that gets it at renewal rather than at the point somebody challenges a decision.
Multi-state scope memo
Where Local Law 144 or Illinois HB 3773 reaches the same tool, and the one piece of testing that satisfies the strictest of them.
Format & scope
Scoped first, then fixed.
- Scope
- The tools that touch a decision
- Data
- Your own applicant outcomes
- Re-test
- On model change - $5,500 against your known baseline
There is no certificate to buy - what is scoped is the testing, and nothing is charged until you approve it.
Start a FEHA bias review
FEHA · Named engagement
$6,500 flat
- Inventory of the tools in scope
- Adverse-impact testing on your own data
- Method and small-category rule stated
- A logged response against every finding
Show your hand
Four things you have to be able to produce
There is no filing, so only the record proves it. Each of these is either in your hand the day someone asks, or not.
The list,
complete
Every tool that takes part in an employment decision, and the ones you considered and then excluded. A system nobody listed is worse than one listed and deferred.
The test,
on you
Adverse impact measured on your own applicants, dated to the model version it describes. Testing on somebody else's population describes somebody else's hiring.
The method,
shown
What was measured, what it was compared against, and how categories too small to speak were handled. A number with no method behind it is not testing at all.
The answer,
logged
Against each finding: remediated, accepted with a stated reason, or open with an owner and a date. This is the part most testing records are missing entirely.
Four cards, and the date on each one is part of the card.
FAQ
Plain answers
Scope, vendor testing, and what cannot be bought. Answered straight.
Start a FEHA bias reviewOur tool only ranks candidates - a person still decides. Are we in scope?
Probably. The question is whether the system takes part in the employment decision, not whether it takes the decision alone. A rank order a human approves without reworking is doing the deciding in practice.
Our vendor published a bias audit. Are we covered?
Only to the extent your own applicants were in it. A vendor audit run on a reference population describes that population - if your data was not in it, you do not have testing about your use of the tool.
Do you file anything for us?
There is nothing to file. That is the shape of this instrument: no submission, no registration and no certificate. What you get is the testing record and the write-up around it.
Can you certify us as FEHA compliant?
No, and neither can anybody else - there is no certification scheme here to be accredited against. Any firm offering you a FEHA ADS certificate is selling you a document with no standing.
We also have roles in New York. Do we need two engagements?
Usually one. Local Law 144 asks for more on its face - an independent annual audit and a published summary - so testing designed to meet that generally carries the FEHA work with it.
Get started
Start a FEHA bias review
Tell us which tools touch a hiring decision and we come back within one business day.
What we need from you
Nothing you do not already have. Most of this comes out of your ATS and your vendor’s paperwork in an afternoon, and we tell you which extracts before you commit.
- Which tools are used, and at which stage of the process
- Whether any roles are in California, or the applicants
- Applicant outcomes, with any demographic fields you hold
- Whether the tool scores, ranks, filters or schedules
- Anything your vendors have given you, however thin
What happens next
- You send the five items we need.
- We call to scope it within one business day.
- Nothing is charged until you approve the scope.
Sources & standing
Where this page gets its facts
Where the claims on this page come from, and what they are worth - stated, not assumed.
What it is drawn from
- California FEHA ADS regulations
- General knowledge of how they are read
- In force
- 1 October 2025
- Made by
- Civil Rights Council
What it means
- General information about what the regulations require - not legal advice, and no professional relationship arises from reading it. It settles nothing about your own tools.
- Where your position is genuinely arguable, our reports say so rather than pick the convenient answer - and they say which way the argument runs, and how far.
Scope & limitation
- Written from general knowledge of the regulations, not checked line by line against them. Do not rest a binding decision on it; engage qualified counsel.
- It covers the FEHA regulations alone - Local Law 144, Illinois HB 3773 and the EU AI Act can reach the same screening tool, and none of them is answered by this one.
- Use it as a starting point for a scoping conversation, not as your final word.
Something on this page out of date?
Tell usFrom Insights
Before you commission one
How to Prepare for an AI Audit: The Readiness Checklist
Six things to have ready before the engagement starts. Assembling them takes a fortnight off the clock — and tends to find the first two findings before an auditor does.
Read the notes
METHODOLOGY NOTES
What Is an AI Audit? Scope, Standards, and What You Get
An independent review of what your AI actually does, measured against a named standard — not a certificate, and not a review of what the documentation says it does.
STANDARDS EXPLAINED
What an AI Governance Framework Actually Contains
Five working parts, not a policy document. What each one has to do, how to tell whether yours is real, and why a framework is not the same thing as compliance.
STANDARDS EXPLAINED
Who Can Run a Local Law 144 Bias Audit, and What It Has to Measure
The law asks for an independent bias audit of the tool you hire with. Two words there do the work — independent, and audit — and a vendor certificate satisfies neither.
