A company that automates part of its hiring process, resume screening, initial ranking, an interview-scheduling tool that also scores candidates, usually treats that as an operations decision: pick a vendor, turn it on, measure the time saved. In New York City, it is also a compliance decision, and the obligation lands on the employer, not the software vendor.
New York City's Local Law 144 regulates the use of an "automated employment decision tool," or AEDT, in hiring and promotion. It is narrow, specific, and already in force: it is not a proposal or a pending rule. Before automating any part of a hiring workflow that touches New York City, four things in the law are worth understanding precisely, because "we use an AI tool" and "we comply with Local Law 144" are not the same statement.
The rule is about audit currency, not a one-time certification
The law does not require an "annual audit" filed once a year on a calendar. It conditions use of the tool on an audit that is current: an employer "may not use or continue to use an AEDT if more than one year has passed since the most recent bias audit," per the New York City Department of Consumer and Worker Protection's final rule adopting the law. That distinction matters operationally. An employer that stops using a tool doesn't owe an audit for a system sitting unused. An employer that keeps using one past the one-year mark is out of compliance the day the audit goes stale, independent of when the calendar year ends.
A summary of the audit has to be public, not filed away
The rule requires publishing "the date of the most recent bias audit of the AEDT and a summary of the results," including the source and explanation of the data used, counts for unknown categories, selection or scoring rates, and impact ratios across categories, "on the employment section of their website in a clear and conspicuous manner." This is not paperwork that satisfies a regulator on request. It is a standing public disclosure, and it has to stay current with whatever audit is actually in effect.
Candidates get advance notice, not a checkbox in an offer letter
Candidates and employees are entitled to notice "at least 10 business days before use of an AEDT." The rule allows that notice to be satisfied through a website posting, a job posting, or direct mail or email: it does not require a specific format, but it does require the notice to land before the tool is used on that candidate, not disclosed after the decision has already been made.
Liability sits with the employer, structurally
Local Law 144 "prohibits employers and employment agencies from using an automated employment decision tool unless" the audit and notice requirements are met. The duty runs to the employer that deploys the tool, not the company that built it. The rule underscores this by disqualifying an auditor from being considered independent if that auditor is tied to "a vendor that developed or distributed the AEDT," treating the vendor as a separate, interested party rather than someone who can vouch for the employer's own compliance. A vendor's marketing claim that its tool is "Local Law 144 compliant" does not transfer the employer's own audit, publication, and notice obligations anywhere.
The compliance clock is longer for employers outside New York, but it is moving
Employers automating hiring outside New York City sometimes treat this as a local curiosity. It is worth knowing that a broader, EU-wide obligation is on a similar track, on a longer clock. Under the EU AI Act, high-risk obligations for standalone systems used in employment and recruitment (human oversight assigned to a qualified person, log retention for at least six months, and advance notice to affected workers and their representatives) were set to apply from 2 August 2026. Regulation (EU) 2026/1744 (the "Digital Omnibus on AI," in force since 27 July 2026) postponed that date to 2 December 2027 for the employment category specifically. An employer building toward that deadline off an outdated date is building toward the wrong one; a plan that already assumed a 2027 timeline was, in this one respect, ahead of where the law actually landed.
What to confirm before automating any part of hiring
None of this requires abandoning automation in hiring. It requires answering a short list of questions before a tool goes live, not after a candidate or a regulator asks:
- Does an independent bias audit exist for this specific tool, and is it less than a year old as of today, not as of when the tool was first adopted?
- Is a summary of that audit published somewhere a candidate could actually find it, in the specific form the rule requires (source and explanation of data, category counts, selection and impact rates)?
- Is the 10-business-day notice built into the hiring workflow itself, so it happens automatically before the tool touches a candidate, rather than depending on someone remembering to send it?
- Who inside the organization owns re-running the audit before it goes stale? Not the vendor, and not "whoever set the tool up originally."
A hiring workflow that can answer all four before deployment is compliant by design. One that can't is running exposure that a vendor's compliance claim does not cover, no matter how the tool is marketed.
If a hiring automation project is on the table and these four questions don't have confident answers yet, that's worth resolving before the tool goes live, not after. Start a conversation about what an AI implementation for hiring actually needs to clear before it ships, or see how this fits into the broader AI governance work that has to exist alongside any automated hiring build.