No Robot Bosses Act
- Sponsor
- Sen. Markey, Edward J. [D-MA]
- Committees
- Health, Education, Labor, and Pensions Committee (primary)
- Last action
- Jun 18, 2026
Bottom line
The 'No Robot Bosses Act' aims to establish comprehensive federal oversight and worker protections against the misuse of artificial intelligence and automated decision systems in employment.
What it actually does
This bill prohibits employers from predominantly relying on automated decision systems for work-related decisions and from using such systems in ways that violate labor or civil rights laws. It mandates extensive disclosures to covered individuals about the use of these systems, requires predeployment evaluations and annual impact assessments by both developers and employers, and establishes a new Fairness and Transparency Office within the Department of Labor to oversee these regulations. The bill also provides a private right of action for affected individuals and prohibits predispute arbitration agreements related to its provisions.
Proponents argue
Supporters argue that the bill is crucial for protecting workers from algorithmic bias, unfair surveillance, and opaque decision-making processes in hiring, management, and termination. They contend it ensures human oversight, promotes transparency, and safeguards fundamental labor and civil rights in an increasingly automated workplace, preventing discrimination and chilling of legally protected activities.
Opponents contend
Critics may contend that the bill imposes overly burdensome compliance requirements on employers and AI developers, potentially stifling innovation and increasing operational costs, especially for small businesses. They might argue that the broad definitions and extensive reporting mandates could create legal uncertainty and make it more difficult for companies to efficiently utilize beneficial AI technologies.
The bill is substantial and detailed, requiring a dedicated reading to grasp its full implications, which is appropriate given its broad scope and potential impact on employment practices.
Section 3(a)(1)
Prohibition on Predominant Reliance and Discriminatory Use
This provision prohibits employers from relying predominantly on automated decision systems when making work-related decisions. It also forbids using such systems in any way that would violate federal, state, or local labor, occupational safety, employment, or civil rights laws, or that would prevent workers from exercising their rights under the National Labor Relations Act or similar laws.
Supporters argue
Supporters argue this provision is essential to ensure that human judgment remains central to employment decisions and to prevent automated systems from perpetuating or creating new forms of discrimination or infringing on worker rights.
Critics contend
Critics might argue that defining 'predominant reliance' could be challenging and lead to ambiguity, potentially discouraging the efficient use of AI tools that could objectively improve hiring and management processes.
Tradeoffs
This provision balances the efficiency and potential objectivity of AI with the need to preserve human oversight and prevent algorithmic discrimination and rights violations.
Section 3(a)(2)
Disclosure Requirements for Automated Decision System Use
Employers must disclose to covered individuals when an automated decision system is used for work-related decisions. This disclosure must include a detailed explanation of the system's design, data inputs, measured characteristics, how outputs are interpreted, the operator's identity, and how the individual can dispute or appeal decisions. Specific timing requirements are set, including advance notice for adverse work actions.
Section 3(b) and 3(c)
Opt-Out Rights for Automated Management and Application Screening
The bill grants covered individuals the right to opt out of being managed by an automated decision system, instead choosing a human manager. Similarly, applicants can opt out of having their application materials screened, reviewed, or processed by an automated decision system, requesting human review instead.
Section 4(a) and 4(b)
Predeployment Evaluations and Annual Impact Assessments
Before deploying or licensing an automated decision system, developers and employers must conduct detailed predeployment evaluations, including design, testing, data use, and potential risks to workers' rights or for discrimination. After deployment, employers must conduct annual impact assessments, reporting on the system's actual effects, data inputs, and any disparate impacts. Developers must also conduct annual reviews of these assessments.
Section 6
Establishment of the Fairness and Transparency Office (FTO)
This section establishes the Fairness and Transparency Office within the Wage and Hour Division of the Department of Labor, headed by a Director appointed by the President. The FTO is tasked with carrying out the Act's duties, including enforcement and rulemaking. It also mandates the creation of several advisory boards (User, Research, Product, Labor) composed of diverse experts and stakeholders to advise the Director.
Section 9(a)(3)
Private Right of Action and Statutory Damages
The bill grants adversely affected covered individuals, labor organizations, and worker advocacy organizations the right to file civil actions against violators of the Act in federal court. Courts may award actual damages, injunctive relief, equitable relief, and statutory damages ranging from $5,000 to $20,000 per violation (or $10,000 to $40,000 for willful/repeated violations of Section 3), and higher for whistleblower retaliation (up to $100,000).
Section 9(f)
Prohibition on Predispute Arbitration Agreements and Joint-Action Waivers
This section declares that predispute arbitration agreements and predispute joint-action waivers are invalid and unenforceable with respect to any alleged violation of the Act. An exception is made for arbitration provisions within collective bargaining agreements between an employer and a labor organization.
Defines 'disparate impact' as an unjustified differential effect on a protected characteristic, and crucially, establishes a presumption that an automated decision system is 'not separable for analysis' unless the developer or employer proves otherwise by a preponderance of the evidence.
Section 2(7)
Why it matters:This is a standard legislative practice for defining key legal terms, but the 'not separable' clause has significant implications for how discrimination cases involving AI would be litigated, making it more challenging for defendants to isolate components of an AI system to argue against a finding of disparate impact.
Case for: Supporters argue this presumption is necessary because complex AI systems often have intertwined components, making it difficult for plaintiffs to pinpoint a single discriminatory element. Analyzing the system holistically ensures accountability for overall discriminatory outcomes.
Case against: Critics might argue this presumption places an undue burden on developers and employers, requiring them to prove a negative or to dissect complex black-box algorithms in ways that are technically difficult or reveal trade secrets.
Estimated impact: Significantly impacts the legal standard and burden of proof for demonstrating algorithmic discrimination in employment, potentially making it easier for plaintiffs to win disparate impact cases.
The definition of 'employer' is exceptionally broad, including any person engaged in commerce employing 11 or more covered individuals, public agencies, the Government Accountability Office, the Library of Congress, and various other federal/state employing offices. It also includes 'any person who acts, directly or indirectly, in the interest of a covered employer'.
Section 2(9)
Why it matters:This is standard legislative practice to ensure broad applicability, but the inclusion of '11 or more covered individuals' and various government entities significantly expands the reach of the Act beyond typical federal labor law thresholds, which often apply to employers with 15 or more employees.
Case for: Proponents argue that a broad definition of 'employer' is necessary to ensure comprehensive protection for all workers, regardless of the size or type of their employer, and to prevent loopholes that could allow entities to evade compliance.
Case against: Critics might argue that extending these complex regulations to employers with as few as 11 individuals, or to numerous government entities, could impose disproportionate compliance burdens on smaller organizations and diverse public sector bodies.
Estimated impact: Vastly expands the number of entities subject to the Act's requirements, including many smaller businesses and a wide array of government offices.
An employer that intends to use an automated decision system for a work-related decision not described in the developer's predeployment evaluation must 'assume the responsibilities of the developer required by this Act' (except for the annual review of assessments).
Section 5(b)(2)(B)
Why it matters:This clause is a critical mechanism for ensuring accountability when employers adapt or repurpose AI systems beyond their intended or evaluated uses. It prevents a loophole where an employer could claim a system was 'off-label' to avoid developer responsibilities.
Case for: Proponents argue this provision is essential to prevent employers from circumventing the Act's requirements by using AI systems in ways not originally evaluated. It ensures that any novel application of AI in employment undergoes the same rigorous scrutiny as a newly developed system.
Case against: Critics might argue that this places an extremely high burden on employers, effectively requiring them to become AI developers and conduct complex technical evaluations for any minor deviation from a system's original scope, potentially stifling innovative internal uses.
Estimated impact: Significantly increases the compliance burden and technical expertise required for employers who wish to customize or expand the use of AI systems beyond their vendor's stated purpose.