A4981 would create a broad regulatory framework for the use of artificial intelligence and related surveillance technologies in employment and in the administration of public benefits and services. It defines and regulates three main categories of systems: automated employment decision systems (AEDS), automated benefit or service decision systems (ABSDS), and electronic monitoring tools (EMT). The bill prohibits a wide range of uses, including monitoring workers in bathrooms or private areas, tracking employees off duty or in personal spaces, requiring monitoring software on personal devices, using biometric or wellness data in employment or benefits decisions, and using AI or surveillance in ways that interfere with labor rights, collective bargaining, privacy, dignity, health, or safety.
Before an AEDS, ABSDS, or EMT could be deployed, the bill would require an impact assessment and a finding of compliance. For employment systems, the assessment would be done by an independent auditor; for public-employee or public-benefit systems, the Department of Labor and Workforce Development would conduct the review. The assessment must examine accuracy, bias, error rates, data sources, intended purpose, and economic effects such as wages, hours, benefits, and potential job losses. Vendors would have to provide extensive technical documentation and data to the auditor or department, and the resulting report would be placed in a public registry with a summary available to affected workers, applicants, service beneficiaries, and their representatives.
The bill also imposes notice, disclosure, access, and appeal requirements. Employers and public entities would have to give advance written notice before deploying or significantly changing these systems, disclose what data is collected and how it is used, and identify any performance standards or productivity quotas. Individuals affected by adverse decisions would be entitled to advance notice, access to relevant records, an explanation of how the system was used, and review by a human decision-maker with authority to modify or overturn the decision. The bill further limits third-party sharing of collected data, requires data accuracy and retention controls, mandates breach notification, and creates civil, administrative, and criminal penalties for violations.
For public employers, the bill adds labor-relations protections. Decisions to enter into automated services agreements or adopt technologies likely to reduce employment would be mandatory subjects of bargaining, and public employers would be barred from entering such agreements during the term of an existing collective bargaining agreement. If public employees are displaced, the bill provides severance, continued health coverage, seniority retention, recall rights, and training for oversight roles. The bill also states that its protections are minimum standards and do not reduce any stronger rights under collective bargaining agreements or other laws.
The overall sentiment reflected in the bill text is strongly protective of workers, unions, privacy, and due process, while still allowing AI and monitoring technologies if they are narrowly tailored, validated, and subject to human oversight. Because no committee transcripts or votes were provided, there is no recorded legislative debate or voting history to indicate broader support or opposition. The main points of contention apparent from the bill itself are the scope of restrictions on employer surveillance and automated decision-making, the extensive disclosure and audit obligations imposed on vendors and employers, and the labor-cost and operational impacts of mandatory bargaining and displacement protections for public employees.
The bill would supplement Title 34 of the Revised Statutes by creating new statewide rules governing AI-driven employment screening, workplace surveillance, and automated public-benefits decisions. It would impose pre-deployment impact assessments, notice and disclosure duties, data-access rights, human-review requirements, retention and security rules, and enforcement mechanisms including administrative penalties, civil liability, and criminal sanctions. It would also affect public-sector labor law by making certain technology decisions mandatory subjects of negotiation and by creating statutory protections for public employees displaced by automation.
The bill’s stated purpose and structure reflect a generally pro-worker, pro-privacy, and pro-union approach to AI governance. It is framed as a guardrail measure rather than a ban, allowing use of these technologies only when they are validated, narrowly used, and subject to human oversight. Because no committee testimony or vote record is available, there is no direct evidence of legislative support or opposition in the provided materials, but the bill itself suggests likely support from labor and privacy advocates and likely concern from employers, public entities, and technology vendors over compliance burdens and limits on automation.
The most notable potential points of contention are the bill’s broad limits on workplace surveillance, its prohibition on relying solely on algorithmic outputs for consequential decisions, and its requirement for detailed impact assessments and public disclosure of assessment summaries. Employers and vendors may object to the cost, administrative burden, and possible exposure of proprietary information, while public employers may object to mandatory bargaining requirements and severance/recall obligations tied to automation-related displacement. Labor organizations and worker advocates are likely to support those provisions, especially the protections against surveillance, retaliation, biased decision-making, and unilateral automation decisions that affect jobs or benefits.