Since July 2023, New York City has required a yearly independent bias audit for automated tools used in hiring and promotion. For two years, almost nothing happened to employers who ignored it. A state audit published in December 2025 explains why, and the city has agreed to change course.
What the audit found
The New York State Comptroller reviewed how the city's Department of Consumer and Worker Protection (DCWP) enforced Local Law 144 between July 2023 and June 2025. The numbers are hard to read any other way:
| In the same 32 companies | Possible violations found |
|---|---|
| The city's own review | 1 |
| The Comptroller's auditors | At least 17 |
In two years, the city received just two complaints about automated hiring tools. The audit found that its complaint system did not reliably send these complaints to the right team, that nobody had checked whether the system worked, and that the city's reviews did not cover everything the law requires. Staff also did not consult the city's own technology office to judge whether a tool counted as an automated employment decision tool.
What the city agreed to do
The Comptroller called for better complaint handling, reviews that cover every requirement of the law, the use of technical experts to evaluate tools, and looking for violations proactively instead of waiting for complaints. The city agreed to implement most of the recommendations, including stronger complaint routing, staff training, written procedures, and interviews and demonstrations of the tools themselves where appropriate.
The penalties have not changed: up to $500 for a first violation and $500 to $1,500 for each one after it, with each day of non-compliance counted separately. What is changing is how likely an employer is to be caught.
What this means for employers
For two years, the realistic risk of using a hiring tool without a valid audit was low. That is no longer a safe assumption. The audit shows that a review can find problems the city missed, and it gives the city a roadmap for finding them.
It also means an audit on file is not automatically a good audit. A bias audit that leaves out a required category, uses the wrong data, or is not published correctly may not hold up when someone finally looks at it closely.
The employers in the best position are the ones that already know what their own tools are doing: which systems count, what the numbers say, and whether the notices and published results match what the law asks for.