After releasing an initial two-page “fact sheet,” Congress publicly posted the bill text of the No Robot Bosses Act (the “Proposed Act”), detailing proposed federal guardrails for use of automated decision-making systems in the employment context. Robert Casey (D-PA), Brian Schatz (D-HI), John Fetterman (D-PA), and Bernie Sanders (I-VT) currently cosponsor the Proposed Act.

The Proposed Act broadly defines Automated Decision Systems (“ADS”) to include systems that use computation, in whole or in part, to determine outcomes, make or aid decisions, inform policy implementation, or collect data or observations. ADS are not limited to systems that use machine learning or artificial intelligence; however, “passive computing infrastructure”—passive systems that do not influence decision-making—would be excluded. 

Employers covered under the Proposed Act include those “engaged in commerce or in any industry or activity affecting commerce who employ, or otherwise engages for the performance of work for remuneration, 11 or more covered individuals[.]” The Proposed Act also includes certain governmental agencies. “Covered individuals” include candidates for employment, as well as individuals “performing work for remuneration,” thus covering both employees and independent contractors.

If enacted, the Proposed Act would subject employers to various requirements when using ADS in connection with “employment-related decisions,” i.e., decisions made during any aspect of the employment lifecycle from candidate selection, interviewing, and hiring, to scheduling, compensation, health benefits, promotion, demotion, and disciplinary action inclusive of termination.

In general, the law would prohibit employers from relying exclusively on an ADS when making employment-related decisions—in effect, requiring some level of human involvement. The law would also require “pre-deployment testing and validation” for ADS with respect to efficacy, compliance with employment discrimination laws, lack of discriminatory impact based on protected categories, and compliance with the NIST AI Risk Management Framework. The discriminatory impact analysis would be required annually with the results made public.

The Proposed Act would require employers that use ADS to provide both an initial disclosure prior to making any employment-related decision and a notice within seven days of making an employment-related decision. The disclosure would be required to include information about the ADS, and how the covered individual can dispute or appeal the employment-related decision.   

The Proposed Act would also, among other things:

• require employers to train individuals who are responsible for operating the ADS, including as to the data that is input, the appeals process, potential biases, potential adverse effects, and potential errors or problems with the ADS; the bill does not, however, define what it means to “operate” an ADS;
• require employers that use ADS to “manage” covered individuals to give the individuals the right to opt out; the bill does not, however, define the term “manage;”
• prohibit retaliation against a covered individual for exercising rights under the Proposed Act, and provide whistleblower protections for individuals who have made a written or oral complaint to the employer or government entity, or otherwise participated in an inquiry or proceeding under the Proposed Act;
• deem any pre-dispute agreements to arbitrate employment-related disputes or pre-dispute waivers of joint, class, or collective actions in any forum invalid or unenforceable with respect to ADS use or whistleblower claims; and
• require employers to independently corroborate an ADS output when making an employment-related decision; the Proposed Act does not state how this independent corroboration must be accomplished, or whether it must be done contemporaneously for each individual employment-related decision.

In terms of enforcement, the Proposed Act would give covered individuals the right to commence a civil action in any Federal court of competent jurisdiction to recover actual or treble damages, injunctive relief, equitable relief, and statutory damages ranging from $5,000 up to $100,000 per violation in certain circumstances. The Proposed Act would further create a Technology and Worker Protection Division in the Department of Labor to investigate and enforce violations, and, if applicable, make criminal referrals for any conduct that “may constitute a violation of Federal criminal law.” The Proposed Act also permits enforcement by State Attorneys General or “State Privacy Regulators,” the latter of which is defined to include both the “chief consumer protection officer of a State” or the “State consumer protection agency with expertise in data protection, including the California Privacy Protection Agency.” 

The proposed No Robot Bosses Act follows a range of activity by the White House and executive branch to intensify the regulation of AI and similar technology. While Congress has not yet passed the Proposed Act, employers should take heed of the direction of these legislative and regulatory efforts. Epstein Becker & Green, P.C. will continue to monitor this and related legislative initiatives and report on them here.