On April 6, 2023, the New York City Department of Consumer and Workforce Protection (“DCWP”) promulgated its final regulations (the “Final Regulations”) regarding the New York City Automated Employment Decision Tools Law (“AEDTL”). In connection with the Final Regulations, the DCWP also notified employers that it would further delay enforcement of the AEDTL from April 15, 2023 to July 5, 2023. The Final Regulations, among other things, expand the definition of “machine learning, statistical modeling, data analytics, or artificial intelligence” as used in the AEDTL and clarify specifics around the bias audits required by the AEDTL.

As we previously reported, the AEDTL’s regulations have remained in a state of flux for months. The Final Regulations are the DCWP’s third attempt to formulate regulations for the AEDTL, and have received an unusually high volume of public comment requiring multiple well-attended public hearings. As a result, enforcement of the AEDTL has been delayed twice; first from January 1, 2023 to April 15, 2023, and now until July 5, 2023. However, because the Final Regulations have been issued by this point, the AEDTL’s enforcement date is unlikely to change further.

Overview of the AEDTL

Once enforced, the AEDTL will restrict employers’ ability to use “automated employment decision tools” in hiring and promotion decisions within New York City. The AEDTL defines “automated employment decision tool” as “any computational process, derived from machine learning, statistical modeling, data analytics, or artificial intelligence, that issues simplified output, including a score, classification, or recommendation, that is used to substantially assist or replace discretionary decision making for making employment decisions that impact natural persons.” The phrase “to substantially assist or replace discretionary decision making” means: (i) to rely solely on a simplified output (score, tag, classification, ranking, etc.), with no other factors considered; (ii) to use a simplified output as one of a set of criteria where the simplified output is weighted more than any other criterion in the set; or (iii) to use a simplified output to overrule conclusions derived from other factors including human decision-making.

Employers may not use automated employment decision tools unless: (i) the tool has been the subject of a bias audit conducted within the previous year in accordance with the AEDTL’s requirements; and (ii) the employer has published a summary of the results of the tool’s most recent bias audit, as well as the distribution date of the tool to which such audit applies, on its publicly-available website. The bias audit must be performed by an “independent auditor,” who cannot have been involved in using, developing or distributing the tool, cannot have an employment relationship with an employer that seeks to use the tool or a vendor that developed or distributed it, and cannot have a direct or material indirect financial interest in the employer that seeks to use the tool or the vendor that distributed it. Employers must use historical data (i.e., data collected during the employer’s use of the tool) to conduct the audit. However, if insufficient historical data is available to conduct a statistically significant audit, employers may use non-historical test data, provided that the employer explains why historical data was not used and how the test data used was generated and obtained.

Under the AEDTL, employers who use automated employment decision tools must also disclose the following information to candidates at least ten business days before the tool is used: (i) the fact that an automated employment decision tool will be used in connection with the assessment or evaluation of any candidate who lives in New York City; (ii) the job qualifications and characteristics that the automated employment decision tool will use in assessing the candidate; and (iii) instructions for how an individual can request an alternative selection process or reasonable accommodation, if available. The AEDTL does not obligate employers to provide an alternative selection process, though employers are otherwise obligated to provide a reasonable accommodation if required under the Americans with Disabilities Act and analogous state and local laws.

Finally, the AEDTL obligates employers to undertake the following additional disclosure steps: (i) provide information in the employment section of its website in a clear and conspicuous manner about its automated employment decision tool data retention policy, the type of data collected for the tool, and the source of the data; (ii) post instructions on the employment section of its website in a clear and conspicuous manner for how to make a written request for such information, and if a written request is received, provide such information within 30 days; and (iii) if such a request is denied, explain why disclosure of such information would violate applicable law or interfere with a law enforcement investigation.

Employers who violate the AEDTL may be subject to civil fines ranging between $500-$1,500 per day that the employer does not comply with the law. The AEDTL neither expressly permits nor prohibits a private right of action, but states that it shall not be construed to “limit any right of any candidate or employee for an employment decision to bring a civil action in any court of competent jurisdiction.”

Impact of the Final Regulations

The Final Regulations imposed several discrete changes to the AEDTL’s requirements and clarified employers’ obligations. More specifically, the Final Regulations:

• Expand the definition of “machine learning, statistical modeling, data analytics, or artificial intelligence” to mean “a group of mathematical, computer-based techniques: (i) that generate a prediction, meaning an expected outcome for an observation, such as an assessment of a candidate’s fit or likelihood of success, or that generate a classification, meaning an assignment of an observation to a group, such as categorizations based on skill sets or aptitude; and (ii) for which a computer at least in part identifies the inputs, the relative importance placed on those inputs, and, if applicable, other parameters for the models in order to improve the accuracy of the prediction or classification”;

• Require bias audits to indicate the number of individuals that the tool assessed that are not included in the calculations because they fall within an unknown category, and requiring that number to be included in the summary of results;

• Permit auditors to exclude categories that comprise less than 2% of the data being used for a bias audit from the calculations of impact ratio;

• Clarify examples of a bias audit;

• Clarify when an employer may rely on a bias audit conducted using the historical data of other employers;

• Provide examples of when an employer may rely on a bias audit conducted with historical data, test data or historical data from other employers; and

• Clarify that the number of applicants in a category and scoring rate of a category, if applicable, must be included in the summary of results.

Next Steps

The AEDTL and the Final Regulations are complex, and this blog provides only an overview. Fortunately, employers who do use or are contemplating the use of automated employment decision tools now have three additional months to comply. In that period, employers should: (i) identify any automated employment decision tools that they currently use and which may be subject to the AEDTL; (ii) begin collecting historical data or, if sufficient historical data is unavailable, identify appropriate test data; (iii) identify an appropriate independent auditor and obtain a bias audit; and (iv) plan to comply with the AEDTL’s notice requirements.