An algorithmic bias audit for one automated tool under New York City Local Law 144 typically costs about $4,950 to $5,500 for the first audit in 2026. That is one of the most concrete price points available, because the law is narrow and its scope is clearly defined. The rule covers automated employment decision tools used in hiring and promotion. It requires an independent bias audit within the past year, public posting of results, and advance notice to candidates. The New York City Department of Consumer and Worker Protection sets out these conditions, so you can check whether your software falls within them before you commit to an audit.
Treat that number as a starting point rather than a final bill. Additional tools commonly add between $900 and $1,500 each, depending on how many you run, and renewals for an unchanged tool generally fall to around $3,950 to $4,500 a year. Turnaround is typically 7 to 10 business days after the provider receives complete, validated data, so the speed of your internal data preparation affects your schedule as much as the price. Before you request a quote, know which tools you use, what each one produces, and who relies on its output. Those answers determine the scope and, with it, the cost you should expect.
Key Takeaways
- For one automated employment decision tool under New York City Local Law 144, you should expect the first bias audit to cost about $4,950 to $5,500 in 2026, while renewals for an unchanged tool typically fall to around $3,950 to $4,500 a year.
- Before you pay for a Local Law 144 bias audit, you should confirm that your hiring or promotion tool is in scope, which applies when its simplified output is used alone, weighted more heavily than other criteria, or used to overrule other factors in the decision.
- Your total audit bill grows with each covered tool, since tools two and three can add $1,500 each on one provider’s tiered schedule, so nine covered tools could total around $15,100.
- Your turnaround depends on data readiness, because providers typically need 7 to 10 business days after receiving a complete, validated dataset, so gathering and cleaning applicant records early protects your hiring calendar.
- Your budget should also account for hidden costs outside the audit fee, including candidate notices that must reach affected applicants at least 10 business days before you use a covered tool, along with data preparation and any fixes the audit findings require.
What Does an Algorithmic Bias Audit Cost in 2026?
For one automated employment decision tool under New York City Local Law 144, published 2026 prices for the first bias audit run from $4,950 to $5,500 at individual audit providers. Renewals are lower. One provider lists annual renewal pricing starting at $3,950, subject to its stated scope, and another lists $4,500 a year for a tool that has not changed. These are general market figures for one tool, so treat them as a baseline rather than a quote.
The first step in reading a quote is knowing what counts as a covered tool. The New York State Comptroller’s December 2025 review explains that a tool is in scope when its simplified output is used alone, when it is weighted more heavily than other criteria, or when it overrules other factors in a hiring or promotion decision. If your software only supplies information that a person weighs alongside everything else, it may fall outside the rule. In that case, you should not pay for an audit you do not need.
Price also rises with the number of tools. One provider’s tiered schedule adds $1,500 each for tools two and three, $1,200 each for tools four through seven, and $900 each for tools eight and above. Complex work, such as unusual data structures or a tool that changes often, can raise the fee further. Providers vary in how they price that work, so ask for each addition to be itemized.
The audit fee covers the analysis itself, not everything your compliance will cost. The law requires notice to affected candidates or employees at least 10 business days before you use a covered tool, and results must be publicly available. Preparing the data, confirming the tool’s inputs and outputs, and having staff handle notices all take time, and none of that appears in a provider’s headline price. If the audit finds problems, fixing the tool or changing how you use it is a separate expense that no audit price includes.
Turnaround also affects your planning. One provider lists five business days from receipt of clean data. Another lists a typical 7 to 10 business days after scope is confirmed and a complete, validated dataset is received, with expedited reviews from five business days subject to capacity and scope. Start the data work early, because no provider can begin until its inputs are complete and accurate.
When a quote arrives, check four things: how many tools it covers, whether the renewal assumes nothing has changed, what triggers an extra charge, and whether data preparation is included. Those answers tell you whether two quotes are actually comparable.
Do 2026 Rules Actually Require an Audit for Your Algorithm?
No, 2026 rules do not broadly require an ethical audit of every corporate algorithm. Whether you must test a system depends on three things: the jurisdiction you operate in, the type of system you run, and the use it is put to. New York City’s Local Law 144 is the clearest concrete example, and it is deliberately narrow, covering only certain automated employment decision tools.
Even that narrow rule is inconsistently enforced. The State Comptroller’s December 2025 review identified at least 17 potential non-compliance instances among 22 employers, while the city’s own review of the same employers identified one. That gap suggests many employers are unsure how the rule applies to their software, which is a good reason to check scope carefully before assuming you are covered or exempt.
Europe takes a different route. The EU AI Act sets duties for specified high-risk areas, including employment, creditworthiness, and certain life and health insurance uses. In its AI Act FAQ, the European Commission explains that providers of high-risk systems must complete conformity assessments before placing them on the market or putting them into service. That is a set of risk-management duties, not one universal ethical audit.
Deployers have their own obligations. Public authorities, and private operators providing public services, must carry out a fundamental rights impact assessment before first use of a high-risk system. Creditworthiness and life and health insurance risk or pricing assessments are named as cases that need one. The high-risk rules have also been extended, with December 2, 2027, for high-risk AI systems and August 2, 2028, for high-risk AI embedded in products.
The third framework you will meet is the National Institute of Standards and Technology’s AI Risk Management Framework. NIST intends it for voluntary use, and it addresses trustworthiness characteristics including fairness, with harmful bias managed. It is a useful reference for building a review process, but it is not a general audit mandate.
For most owners, the practical question is narrower than whether ethical audits are required at all. Ask where your applicants or customers are located, whether any tool feeds hiring, promotion, credit or insurance decisions, and whether a person reviews its output in a way that changes the result. Those answers show which framework applies and whether a bias audit is a legal duty or a voluntary safeguard.
What Drives the Price of a Local Law 144 Bias Audit?
Four things move the price of a New York style bias audit: how many tools you run, which categories the audit must test, how clean your data is, and how quickly you need results. Tool count usually has the largest effect, and a typical tiered schedule shows how quickly the total grows. Suppose you run nine covered tools. The first audit costs $5,500, tools two and three add $3,000 together, tools four through seven add $4,800, and tools eight and nine add $1,800. That brings the total to $15,100, close to three times the single-tool price.
The categories tested also shape the work. Under Local Law 144, audits must analyze sex, race and ethnicity, and intersectional categories. Intersectional analysis looks at combinations of categories, such as race and sex together, rather than each one alone. Each added combination means more groups to compare. A tool that scores applicants for several roles or locations can therefore require more analysis than a simple one, even when the software looks similar on the surface.
Data quality is the factor you control most directly, and it is the one that most often changes a timeline. Messy or incomplete data is a common reason schedules and scopes shift after work begins. Typical problems include applicant records with missing demographic fields, duplicate candidate entries, and tool outputs that were never saved for some hiring stages. One provider lists a typical turnaround of 7 to 10 business days after scope confirmation and receipt of a complete, validated dataset. Another lists five business days from receipt of clean data. In both cases the clock starts only once the data is ready, so gaps can delay the audit and may also reopen the scope.
Turnaround is also a price question. Some providers offer expedited reviews from five business days, but those depend on capacity and scope, so a rush request is not guaranteed. If your hiring calendar is tight, ask early whether a faster slot is available for your tool count and categories.
Before you request a quote, assemble a short inventory: every covered tool, the applicant fields each one uses, where its outputs are stored, and which hiring stage each output affects. Providers can price a scope they can see, and an inventory like this makes the quotes you receive far easier to compare.
Where Do the Hidden Costs of an Audit Sit?
The hidden costs of an audit sit mostly in your own staff time and in the work around the audit, not on the provider’s invoice. They fall into four areas: notices to candidates, data preparation, internal coordination, and what you do once the findings arrive.
Start with notice. Under Local Law 144, you must tell affected applicants or employees at least 10 business days before you use a covered tool. Someone has to identify who is affected, write the notice, send it on time, and keep a record that it went out. If your hiring calendar is already tight, that window can push back interviews that were scheduled months ago.
Data preparation is often the largest hidden cost. Before an auditor can start, you may need to gather historical outcomes from past hiring rounds, confirm that applicant records have complete demographic fields, and document how the tool works: what inputs it reads, what score or recommendation it produces, and who acts on that output. Most of this work falls to recruiters and HR administrators, not compliance staff, and their hours rarely appear on any quote. Across a busy quarter, that time can rival the audit fee itself.
Internal coordination adds more. Someone must decide who speaks for the company, answer the auditor’s follow-up questions, and confirm that the tool’s description matches how managers actually use it. If a recruiter regularly overrides a score, the audit scope may change, and that conversation needs to happen before work begins.
Legal review and public posting are separate tasks, and a budget for one does not cover the other. Counsel will want to read the results before anything is published. Posting the summary is then its own step, involving formatting, approval, and placing it where candidates can find it.
The enforcement picture makes the cheapest option risky. Given the gaps between compliance and review, a light review does not prove you are safe. A cheap audit that covers the wrong tool, skips an intersectional category, or relies on incomplete data can leave you exposed. Fixing that scope in a second round, or answering a complaint, can cost more than a properly scoped audit would have at the start.
Remediation comes last and is often left out of budgets. If findings show uneven outcomes, you may need to adjust the model, change how a person reviews its output, or stop using the tool for certain roles. Each option takes staff time and may slow hiring, so plan for that possibility before the results arrive.
How Should You Budget for Repeat Audits and EU Assessments?
Budget for New York City audits every year, not once. Local Law 144 requires a bias audit within one year before you keep using a covered tool, so the first audit starts a recurring cost. Renewal for an unchanged tool costs less than the first audit at both providers, so your later-year figure should be lower. That saving only holds if the tool has not changed. If you alter its inputs, retrain its model, or start using its output for a new role, the provider may treat the work as a new scope with a higher fee. Ask each provider in advance what counts as a change, so the next invoice does not surprise you.
The EU route is budgeted on a different schedule. Under the EU AI Act, providers of high-risk systems must complete a conformity assessment before they place the system on the market or put it into service. If you buy such a system rather than build it, ask the vendor whether that assessment is complete before you sign, because the answer determines whether you face extra work.
Deployers have their own duties. Public authorities, and certain private operators that provide public services, must carry out a fundamental rights impact assessment before first use of a high-risk system. Creditworthiness assessments and life and health insurance pricing assessments are named cases that need one. If your business lends money or prices insurance, plan for that assessment as a task completed before launch, and set aside time to revisit it whenever the system changes.
The cost of these assessments depends on three factors. The first is system complexity: a tool with many data sources or one that feeds several decisions takes longer to document and test. The second is internal staff time, since a team that already monitors model performance will need fewer hours than one starting from scratch. The third is whether outside experts are involved. Specialists add fees, but they can speed up work your staff lack the time or skills to finish. Ask for estimates that separate internal hours from external fees, so you can see which part of the bill you control.
Scoping Your Audit Before You Ask for a Quote
Start by listing every automated tool that touches a hiring or promotion decision. For each one, record whether its output is used alone, weighted more heavily than other criteria, or used to overrule other factors, because the New York City rules treat all three situations as in scope. A resume screener that ranks applicants and passes only the top ten to a recruiter is a typical example of a tool whose ranking is used alone. Then confirm whether the rule applies at all. Check where your candidates are located, whether the role is based in New York City, and whether the tool feeds a covered decision. If the answer is no, you may not need a Local Law 144 audit, and you should not pay to test a tool that falls outside the rule.
Once you know which tools are covered, prepare a complete, validated dataset before you contact any provider. Fill in the missing demographic fields in applicant records, remove duplicate candidate entries, and confirm that the output of every hiring stage has been saved. Add a one-page note describing what each tool reads as input, what it produces, and which person acts on that result. Providers cannot start until this material is in order, so a clean file is the fastest way to protect your schedule.
When you request quotes, ask every provider to answer in the same terms. Each quote should state the number of tools covered, the categories tested, including sex, race and ethnicity, and intersectional combinations, the renewal price for an unchanged tool, and the turnaround measured from the day your data is accepted. Quotes that answer those points can be compared line by line. Anything left vague is a question to settle before you sign.
Getting a Fair Quote Without Guessing at Scope
Your next step is a single spreadsheet. List every tool that touches a hiring or promotion decision, one row per tool. For each one, note the hiring stage it affects, whether its output is used alone, weighted more heavily than other criteria, or used to overrule other factors, and the person who acts on that output. That table shows you which tools need a quote and which you can set aside.
Then count backward from your next hiring date. Affected candidates must receive notice at least 10 business days before you use a covered tool, and a provider typically needs 7 to 10 business days after accepting your validated data. If you plan to use a tool in six weeks, the notice deadline and the data deadline may already be close, so mark both on the same calendar.
Finally, send the same written questions to at least two providers. Ask how many tools the quote covers, which categories are tested, what an unchanged renewal costs, what counts as a change that triggers a new fee, and whether the turnaround starts only once your data is accepted. Request the answers in writing. If a provider cannot answer one of those questions clearly, you have found the point to settle before you sign.
Frequently Asked Questions
1. Can my own team run a bias audit to save money?
Local Law 144 calls for an independent bias audit, so a review run only by your own staff will not, on its own, meet the requirement. Internal checks can still help you spot uneven outcomes early and prepare clean data for the outside auditor. Confirm the scope against the rule before you cut the external audit to save money, since skipping a required audit can leave you exposed.
2. What is a fundamental rights impact assessment in plain terms?
A fundamental rights impact assessment is a structured review of how a high-risk AI system could affect people’s basic rights. Under the EU AI Act, public authorities and private operators that provide public services must complete one before first use of a high-risk system. Creditworthiness assessments and life and health insurance pricing assessments are named cases that need one, and the assessment should be revisited whenever the system changes.
3. What is an AI bias audit?
An AI bias audit is a review that tests whether an automated system produces different results for different groups of people. Under New York City’s rules for hiring tools, the audit must analyze sex, race and ethnicity, and intersectional combinations of those categories. The results are then made publicly available so candidates can see how the tool performed.
