In the rapidly developing world of AI, federal regulators are again signaling that businesses and HR managers cannot depend on a “data made me do it” defense to employment decisions made with the help of AI systems. Expanding on guidance it released in May, the US Department of Labor (DOL) issued new guidance on October 16th reminding employers they cannot hide behind an algorithm when an AI-generated employment decision runs afoul of federal law.

Although the guidance states it is not creating any additional rules, it reminds employers that existing laws still apply to their use of AI. In other words, if you couldn’t do it without AI, you still can’t do it with AI. Though this sounds simple enough, ensuring compliance with existing laws can be tricky for most employers, which have little to no control over the AI platforms they use. Given this, employers need to know what responsibilities (the DOL believes) they have when using AI in the workplace.  

The DOL’s guidance centers around eight high level principles first announced in President Biden’s 2023 Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. The guidance is heavy on its use of aspirational language, but here is a summary of the practical advice:

  1. Centering Worker Empowerment – The DOL encourages employers to involve workers (particularly those in underserved communities) in AI design, development, testing, training, procurement, deployment, use, and oversight. Borrowing language from the traditional labor context, the guidance also directs employers to “bargain in good faith” with labor unions on the use of AI (particularly monitoring) in the workplace.
  2. Ethically Developing AI – This portion of the guidance focuses on workers’ civil rights, safety, and job quality. Some pitfalls against which the DOL cautions include AI systems with high error rates or systems that evaluate workers based on discriminatory performance standards. To combat these issues, the DOL encourages AI developers to create jobs dedicated to training and refining AI systems.
  3. Establishing AI Governance and Human Oversight – As its title suggests, this segment encourages employers to regularly evaluate and refine their AI systems. The DOL recommends that employers: (a) adopt policies regulating the implementation and use of AI; (b) provide training on AI systems, including on how to interpret AI recommendations; (c) limit the role of AI in making “significant employment decisions”; (d) have an appeal procedure to challenge AI recommendations; and (e) ensure these systems are regularly audited.
  4. Ensuring Transparency in AI Use – The guidance recommends providing workers prior notice about employers’ use of AI systems, what data they collect, and how that data is used. It also encourages employers to allow employees to dispute AI decisions and submit proposed corrections without retaliation for making those reports.
  5. Protecting Labor and Employment Rights – Employers are reminded that their AI-generated decisions are just as much subject to employment laws as are decisions made by HR professionals and hiring managers. For example, employers using AI in their hiring processes—and particularly those using third-party platforms for AI decisioning—should ensure their systems do not carry forward discriminatory assumptions or have a disparate impact on protected groups of candidates.
  6. Using AI To Enable Workers – Referencing its Good Job Principles, the DOL generally encourages employers to use AI in a way that improves job quality for their employees. Practical suggestions include: piloting AI systems before putting them into use, minimizing employee monitoring, and using AI to enhance the predictability of workflows and scheduling.
  7. Supporting Workers Impacted by AI – The DOL acknowledges that introducing AI in the workplace means that some jobs may become obsolete. In response, the DOL encourages employers to uptrain workers who are likely to be displaced by AI and to find other positions for them within the organization when feasible.
  8. Ensuring Responsible Use of Worker Data – Finally, the guidance emphasizes the need to protect worker privacy, stressing that employers should not collect more data about employees than is necessary to make legitimate employment decisions and safeguard employee data from unauthorized access. Although the DOL’s guidance does not carry the force of law, it provides insight on how the DOL may enforce current law involving the use of AI in the workplace. And, depending on the results of the election, it also may foreshadow what regulations the agency may propose in the future. We will continue to monitor any developments, but in the meantime, remember the DOL isn’t going to buy a “the data made me do it” defense.

Although the DOL’s guidance does not carry the force of law, it provides insight on how the DOL may enforce current law involving the use of AI in the workplace. And, depending on the results of the election, it also may foreshadow what regulations the agency may propose in the future. We will continue to monitor any developments, but in the meantime, remember the DOL isn’t going to buy a “the data made me do it” defense.