New York City is reviewing a proposed artificial intelligence law

in 2021, New York City has enacted a measure banning the use of automated hiring decision tools (“AEDT”) to (i) screen job candidates, or (ii) assess current employees for promotion, unless the tool was subject to a “bias audit, conducted at least one year prior to use of the tool.” The law also requires certain notifications to be made. related to the use of ATMs by job seekers. The measure, known as Local Law 144 of 2021, was due to go into effect on January 1, 2023.

In September 2022, the New York City Department of Consumer and Workers Protection (DCWP) released guidance on the new law and announced that it was hosting a public preliminary hearing. After the hearing, the DWCP announced that the law would not be implemented until April 1, 2023, due to the large number of public comments it had received in response to previous hearings.

At the end of December 2022, the DCWP was released Revised proposed rules To implement the decree and set another date Public hearing on January 23, 2023. These proposed rules modify the initial proposed rules. The comment period on proposed regulations will remain open until January 16, 2023.

Here are some important highlights of the recently released rules:

Automated Employment Decision Tools Definition (AEDT) Amendment

By the decree, an AEDT is defined as any computational process, derived from machine learning, statistical modeling, data analytics, or artificial intelligence, that produces a simplified output, including a score, rating, or recommendation, that is used to significantly assist or replace Make discretionary decisions Employment decisions affecting natural persons.

The recent proposed rules seek to clarify this definition by noting that “to substantially assist or replace discretionary decision-making” means:

(1) relying only on simplified outputs (score, mark, rating, ranking, etc.), not taking into account any other factors;

(ii) to use a simplified output as one of the set of criteria where the simplified output is weighted more than any other criterion in the set; or

(3) to use simplified outputs to invalidate conclusions drawn from other factors including human decision-making.

Clarification regarding bias checks

The proposed rules also aim to clarify the meaning and scope of bias audits and independent reviewers.

Bias checks The proposed rules indicate that historical data can be used to perform a bias check. Notably, if there is insufficient historical data to perform a statistically sound bias check, test data can be used. But if test data is used, the summary of the bias audit required should state the reason(s) for not using historical data and describe how the test data used was generated and obtained. And if multiple employers use the same bias control tool, they may rely on the same bias audit so long as they provide historical data, if any, for the independent auditor to consider in that bias audit. Employers must ensure that they rely on bias audits that are no more than one year old.

Independent auditors – The proposed rules also seek to end any uncertainty as to what constitutes an “independent auditor”. Under the new definition, an “independent auditor” may not be employed or have a financial interest in an employer or employment agency that seeks to use or continue to use the AEDT or in a vendor that has developed or distributed an AEDT instrument.

It is understood that these changes represent only a fraction of the proposed rules that will be discussed at the upcoming hearing.