Hiring teams at early-stage startups face a growing stack of AI compliance obligations, and most don’t have a dedicated legal or HR function to track them. NYC Local Law 144 set the opening standard: any employer using automated tools to screen candidates in New York City must conduct and publish an annual bias audit or stop using those tools. Since then, Colorado, Connecticut, Illinois, and California have each passed their own AI hiring laws, with deadlines arriving through 2027. This guide covers what NYC Law 144 requires, how the 2026 compliance wave changes the picture, and what to do about it.
TLDR:
-
NYC Law 144 bans AI hiring tools without a published bias audit; violations cost up to $1,500 per day per tool.
-
Bias audits must measure selection rates and impact ratios across sex, race, ethnicity, and intersectional categories annually.
-
Publish each tool’s audit results at least 10 business days before use, and pause any tool without a completed audit.
-
Connecticut, California, Illinois, and Colorado have all passed their own AI hiring laws with deadlines arriving through 2027.
-
Some tools run annual bias audits for all customers, including free-tier users, covering both the ATS and recruiting layers.
What Is NYC Law 144?
NYC Law 144 requires companies to limit the use of AI-powered employment tools to those that have published results of a customized bias audit.
The law applies to any HR tool that uses algorithms, machine learning, or AI to sort or assess applicants, known as Automated Employment Decision Tools (AEDTs).
Under Local Law 144, employers and employment agencies are prohibited from using AEDTs for employment decisions without conducting annual bias audits, publishing public summaries of the audits, and providing specific notices to applicants and employees subject to the screening.
Violations will result in fines up to $1500 per day for each tool that’s non-compliant.
What Is a Bias Audit?
The goal of a “bias audit” is to understand whether or not AEDTs treat people differently based on their race, ethnicity, or sex.
That means you’ll have to hire an independent auditor to check if the AEDT’s rules cause unfairness to certain groups of people. The law doesn’t clarify how the audit should be done or what counts as passing or failing. But it does say that the audit needs to happen every year by someone who is not involved with the company, and a summary of the results needs to be posted on the company’s website.
Bias audits review the selection rates and impact ratios across a defined set of categories to check for how they’re distributed throughout your process.
-
Selection rates measure how often candidates in each category are selected to move forward. The rate can be calculated by dividing the number of individuals in the category moving forward or assigned a classification by the total number of individuals in the category who applied for a position or were considered for promotion.
-
Impact ratios measure either (1) the selection rate for a category divided by the selection rate of the most selected category or (2) the scoring rate for a category divided by the scoring rate for the highest scoring category.
In order to be compliant, you need to measure selection rates and impact ratios across these groups:
| Category Type | Example Comparison | Measured By |
|---|---|---|
| Sex | Male candidates vs. female candidates | Selection rate and impact ratio per sex group |
| Race / Ethnicity | Hispanic or Latino vs. Black or African American (Not Hispanic or Latino) | Selection rate and impact ratio per race/ethnicity group |
| Intersectional (sex + race/ethnicity) | Hispanic or Latino male vs. Black or African American (Not Hispanic or Latino) female | Selection rate and impact ratio per intersectional group |
The tool’s end results, referred to as “simplified outputs”, like scores, rankings or recommendations, are also reviewed. Bias Audits must be conducted within 1 year of using the AI-assisted tools.
What You Need to Do to Comply
To comply with the law, you must:
-
Publish the results of a Bias Audit for each AI employment tool at least 10 business days before using it, and repeat annually. As a reminder, Dover conducts annual bias audits for all customers, including those on free plans.
-
Pause use of any tools without a published audit until one is completed.
-
Ask tool vendors and software providers about compliance and auditing as you review new AI recruiting tools
For teams managing multiple AI hiring tools across different parts of the funnel, tracking compliance across each vendor is a coordination task in its own right. A fractional recruiting agency with startup hiring experience can take on that vendor-vetting work: reviewing which tools have published bias audits, flagging gaps before a role goes live, and working alongside legal or HR advisors as new state-level deadlines arrive. For most seed-to-Series B teams without dedicated HR or legal resources, this kind of cross-tool compliance work tends to fall through the cracks between the ATS and the legal team, and an agency partner can own it without requiring a full-time HR hire.
The 2026 AI Hiring Compliance Wave
NYC Law 144 was the opening move. What’s arrived in 2026 is a coordinated series of state-level deadlines that affect nearly every company hiring remotely in the US.
The most immediate: Connecticut’s CART Act takes effect October 1, 2026. It imposes obligations on developers and deployers of automated employment-related decision processes, amends the Connecticut Fair Employment Practices Act to clarify that using AI is not a defense to a discrimination claim, and requires employers filing WARN Act layoff notices to disclose whether AI contributed to those layoffs. Core obligations phase in through October 1, 2027.
California’s AI Transparency Act took effect August 2, 2026, requiring deployers of AI systems used in hiring, promotion, and related employment contexts to give applicants notice and meet disclosure requirements borrowed directly from the EU AI Act’s Article 26 framework.
Illinois HB 3773, in effect since January 1, 2026, prohibits employers from using AI that causes discriminatory effects across the full employment lifecycle, including hiring, promotion, discharge, and more, and bars using ZIP code as a proxy for protected class. The Illinois Department of Human Rights has released draft regulations covering resume screeners, video interview analyzers, personality assessments, and AI analysis of third-party candidate data. Employers must notify workers when AI is used in a covered decision.
Colorado’s replacement bill, SB 26-189 (signed May 2026), repeals the original AI Act and replaces it with a narrower framework effective January 1, 2027. The new law targets automated decision-making technology used in consequential employment decisions and is enforced by the state Attorney General as a deceptive trade practice.
At the federal level, the EEOC has affirmed that AI tools fall within Title VII enforcement and that the four-fifths (adverse impact) rule applies to algorithmic screening. Long-standing civil rights statutes apply to AI-driven employment decisions regardless of whether new legislation passes.
The EU AI Act classifies employment AI as high-risk. Deployer obligations for employment-related systems were originally set for August 2026, but a 2026 amendment has pushed that deadline to December 2, 2027 for organizations with EU workforce exposure.
For companies hiring across jurisdictions, compliance is no longer a single audit. It’s an ongoing calendar of annual bias audits, applicant notices, and disclosure requirements, each with its own timeline, scope, and enforcement mechanism.
How to Evaluate AI Hiring Tools for Compliance
With bias audit requirements now spanning multiple states and the EU, the practical question for most startup teams is not whether to comply but how to confirm that each tool in their stack already does. Vendors vary widely in how they approach this, and the burden of verification falls on the employer.
Start with a direct question to each vendor: has an independent bias audit been completed? Ask for the published summary, and do not accept only a verbal confirmation. A compliant audit under NYC Law 144 will include selection rates and impact ratios broken down by sex, race/ethnicity, and intersectional categories, conducted by an auditor with no financial stake in the tool. If a vendor cannot produce a published summary, that tool cannot be used for NYC candidates without putting the employer in violation.
When reviewing a published audit, check the date. Audits must be renewed annually, so a report from two or more years ago is no longer current. Also check that the audit covers the specific tool or feature you are using. Some vendors audit one product but not others in their suite, and an audit of the core ATS does not cover an add-on resume screener or video interview analyzer if those are separate modules.
For teams managing several AI hiring tools across sourcing, screening, and scheduling, tracking compliance across each vendor is a coordination task in its own right. A fractional recruiting agency with startup hiring experience can take on that vendor-vetting work: reviewing which tools have published audits, flagging gaps before a role goes live, and working alongside legal or HR advisors as new state-level deadlines arrive. Dover’s free ATS gives teams a shared pipeline so internal hiring managers and fractional recruiters work from the same system, with verified reviews available for each recruiter in the marketplace before a search begins.
Frequently Asked Questions
Does NYC Local Law 144 apply to early-stage startups and YC-backed companies?
Yes. The law applies to any employer or employment agency using an Automated Employment Decision Tool (AEDT) to screen candidates for roles in New York City, regardless of company size or funding stage. Early-stage and YC-backed startups that use AI-powered resume screeners, automated scoring tools, or interview analysis software to hire in NYC must either conduct an annual bias audit with a published summary, or stop using those tools for NYC-based decisions. There is no startup exemption. If you are using an off-the-shelf AI hiring tool, check with the vendor about their bias audit status before using it for any NYC candidate.
Which AI hiring tools commonly used by startups require a bias audit under NYC Law 144?
Any tool that uses machine learning or algorithms to rank, score, or filter candidates counts as an AEDT under the law. Common examples include AI resume screeners, automated phone or video interview analyzers, personality and skills assessments that generate scores, and ATS features that auto-rank applicants. If you are using a tool like this and have candidates in NYC, the vendor should have a published bias audit. Ask for it before you go live.
Do I need a recruiter if I’m already using an ATS for NYC Law 144 compliance?
An ATS handles the mechanics of your pipeline: candidate tracking, stage management, and job posting distribution. It does not, on its own, audit or certify the AI tools you use to screen or score applicants. Under NYC Law 144 and the broader wave of state-level AI hiring laws, the compliance obligation sits with the employer. You need to confirm that every AEDT in your stack has a published bias audit, and not simply the tool you use most often. For teams without dedicated HR or legal resources, a fractional recruiting agency with startup experience can manage that vendor-review process on an ongoing basis: checking which tools have published audits, flagging gaps before you use a tool for NYC candidates, and keeping pace with new state deadlines as they arrive. Unlike a one-off recruiter hire, an agency partner builds institutional knowledge of your stack across searches. Dover’s free ATS runs its own annual bias audits for all customers, and agency recruiters working through Dover operate within that same shared system, with verified reviews available for each recruiter in the marketplace.
Final Thoughts on AI Hiring Compliance and the Obligations Ahead
NYC Law 144 was the first binding requirement to put an annual bias audit between an employer and its AI hiring tools. What has followed, including Connecticut, Illinois, Colorado, California, and the EU AI Act, confirms that the audit-and-disclosure model is becoming the baseline for any company hiring across jurisdictions. The compliance calendar is no longer a one-time check; it is a recurring obligation tied to specific deadlines, each with its own scope and enforcement mechanism. Dover’s free ATS runs annual bias audits for all customers and fractional recruiters working through Dover operate within that same audited system, which means teams using Dover have one fewer coordination gap to close as new state-level deadlines arrive. For teams still sorting out which tools in their stack have published audits and which do not, the framework here is the starting point.
