New York lawmakers want businesses to report how AI affects layoffs, hiring, work hours and unfilled jobs. The harder question is how employers determine which workforce changes AI actually caused.
getty
Artificial intelligence is taking people’s jobs. At least, that is what a growing number of layoff announcements seem to suggest.
U.S. employers cited AI in connection with 10,970 announced job cuts in July, according to Challenger, Gray & Christmas. AI was the leading reason for announced cuts for the fifth consecutive month. Through June, employers had attributed more than 100,000 planned job cuts to AI in 2026.
Those numbers appear to tell a clear story about AI’s impact on employment. Look more closely, however, and a harder question emerges: How do we actually know when AI caused someone to lose a job?
New York is trying to find out.
Earlier this year, New York lawmakers passed legislation that would require many businesses to report annually on how AI affected their workforces. Assembly Bill A9581B passed both chambers in June and, as of this writing, has not yet been delivered to Gov. Kathy Hochul.
If signed, the legislation could create one of the more ambitious state efforts to measure what AI is actually doing to jobs. It could also reveal just how difficult that is to quantify.
New York Has Already Started Counting
The effort did not begin with A9581B.
In her 2025 State of the State agenda, Hochul directed the New York Department of Labor to require businesses submitting Worker Adjustment and Retraining Notification, or WARN, notices to disclose whether a layoff was related to the employer’s use of AI. The administration described the initiative as a way to understand the effects of new technology through “real data.”
New York’s WARN Act generally requires covered businesses to provide 90 days’ notice before certain plant closings, mass layoffs, relocations and reductions in work hours. Adding AI to the state’s WARN reporting system created a relatively straightforward experiment: When employers conduct covered layoffs, ask whether AI played a role.
The early results are revealing.
New York’s 2026 WARN data currently include one employment action that expressly identifies artificial intelligence as a reason. Nespresso filed a WARN notice concerning an action affecting 46 workers at its New York City location. The Department of Labor lists the reasons as “Relocation of Business, Artificial Intelligence.”
The entry raises as many questions as it answers. The public record does not indicate how much of the employment action resulted from relocation and how much resulted from AI. Nor does it explain whether AI directly replaced particular employees, enabled a broader restructuring or contributed in some other way.
Workforce decisions rarely have a single cause, which makes counting AI-related job losses complicated.
A Much Bigger Experiment
A9581B would take New York’s effort well beyond layoffs covered by WARN.
The bill applies to businesses doing business in New York that employ more than 50 people, as well as publicly traded businesses. An earlier version set the employee threshold at more than 100, but lawmakers expanded the bill’s coverage before passing it.
Covered businesses would have to report to the Department of Labor by March 1 each year about AI’s impact during the preceding calendar year.
The required employment data would include estimates of the number of employees displaced or whose hours were reduced because of AI, employees hired or whose hours increased because of AI, and previously filled positions the business decided not to fill because of AI.
The bill repeatedly uses an important phrase: “due in full or in part” to the use of artificial intelligence.
Employers would also report information about how they use AI, including its objectives, human oversight, frequency and duration of use, use involving sensitive personal data, and measures for oversight and risk reduction. The Department of Labor would aggregate the submissions and publish an annual report analyzing AI’s employment effects by sector, geography and business size. Businesses that fail to report could face civil penalties of up to $500 per day.
AI Can Change A Workforce Without A Layoff
WARN notices capture only one way technology can affect employment.
Suppose an employee leaves a company voluntarily and AI allows the remaining team to absorb that person’s work. The company decides not to replace the departing employee. No one was laid off, and the departure itself would not ordinarily trigger WARN.
Yet a position that once existed has disappeared. Did AI eliminate a job?
Similar effects could appear in other ways. An employer might reduce hiring, reallocate work among existing employees or create new positions to implement and oversee AI. Employees’ duties and work hours may also change.
None of those developments necessarily produces a mass layoff.
A9581B appears designed to capture these less visible workforce changes. Its sponsor memorandum acknowledges uncertainty about whether AI-related displacement will occur through direct layoffs or through reduced hiring as workers leave their jobs.
Much of the policy focus on workplace AI has centered on how employers use algorithmic tools to make decisions about applicants and employees. A9581B takes a different approach by asking employers to quantify how AI affected workforce size and composition after deployment.
The challenge, however, is determining when and to what extent those changes occurred because of AI.
When Does AI “Cause” A Job To Disappear?
Consider a company that introduces generative AI across its operations. Productivity improves. At the same time, economic conditions weaken and management launches a cost-cutting initiative.
Over the next year, 100 employees leave. The company fills only 70 of those positions because managers conclude that existing employees, assisted by AI, can handle some of the work.
How many positions were left unfilled because of AI?
Perhaps 30. Some portion of those positions might have disappeared anyway because of the company’s financial condition. AI might also have been one factor in a larger restructuring decision.
A9581B anticipates some of that complexity by asking employers for estimates and capturing effects caused “in full or in part” by AI. The broader inquiry also introduces considerable judgment into the resulting data.
Two companies experiencing similar workforce changes could reach different conclusions about whether AI contributed to them. One might attribute an unfilled position to AI because automation allowed other employees to absorb the work. Another might characterize the same decision as ordinary attrition, restructuring or cost reduction. How businesses characterize those decisions could become another variable in the data.
The methodology the Department of Labor eventually develops could therefore matter almost as much as the reporting obligation itself.
The bill directs the department to create standardized reporting forms and processes and allows it to develop additional reporting requirements. Clear definitions and consistent methodologies could determine whether the resulting statewide data provide a meaningful picture of AI’s workforce impact.
What Exactly Are We Counting?
The difficulty is already visible in national data.
Challenger reported that employers cited AI in connection with 54,836 announced job cuts in 2025. By the end of June 2026, that number had already reached 101,743 for the year. In July, employers attributed another 10,970 announced cuts to AI.
Even Challenger has cautioned against treating those figures as a precise measurement of causation. Earlier this year, the firm noted the difficulty of determining AI’s specific impact on layoffs while companies increasingly discuss AI implementation and markets respond to those announcements.
Counting the number of job cuts that companies associate with AI is one measurement. Determining how many jobs would have existed without AI is a much harder one.
New York’s proposal attempts to get closer to the second question by capturing layoffs, hiring, hours and vacancies directly from employers. Yet employers will still have to make the underlying causal judgment.
The state’s eventual statistics will therefore depend on thousands of individual decisions about what counts as an AI-related workforce change.
The Next Phase Of AI Employment Regulation
New York’s approach reflects a broader evolution in the AI policy debate.
Hochul has pursued AI investment and workforce adoption while also emphasizing the technology’s potential disruption to workers. In April, she described AI as a major shift in the labor market while announcing a state initiative focused on its effects on workers. The state has also expanded AI training to more than 100,000 state employees.
A9581B fits within that effort by trying to create something policymakers currently lack: a recurring dataset showing how businesses say AI is changing employment.
If the bill becomes law and the reporting system works as intended, New York could eventually offer a more detailed view of AI’s workforce effects than traditional layoff statistics provide. Policymakers could see whether AI-related displacement clusters in particular industries or regions, whether companies create jobs alongside those they eliminate, and whether the larger employment effect appears through layoffs or through positions quietly left unfilled.
Those data could shape future debates over worker retraining, education, economic development and AI regulation.
Their usefulness will depend on whether employers can consistently answer the question at the center of the entire exercise.
AI may be transforming work. Measuring that transformation requires deciding when AI actually caused a job to disappear, appear or change.
That sounds straightforward until someone has to decide what counts.