AI-Driven Layoffs in Canada: Employer Compliance Guide

Canadian employers may use technology to support workforce planning, but they remain responsible for every employment decision. An AI recommendation, productivity score or automated ranking should never be accepted without meaningful human review. Artificial intelligence is changing how employers recruit, monitor and evaluate workers. But should an AI system help decide who loses their job? That question gained attention after 26 Meta employees filed a U.S. lawsuit alleging that AI-supported productivity and activity measurements unfairly affected workers who had taken medical, parental, disability or family-related leave. The allegations have not been proven. Meta denies that AI made the layoff decisions and says people made the final selections. Still, the case offers an important warning for Canadian employers using automated tools in restructuring and termination decisions. According to the complaint, the 26 plaintiffs were still employed when the case was filed, with their separations scheduled to begin on July 22, 2026. The lawsuit was filed in the United States and is based on American law. Canadian employers are subject to different legislation. However, a similar situation in Canada could raise concerns under: Federal, provincial or territorial human rights legislation Employment standards and job-protected leave requirements Privacy legislation and employee-monitoring rules Disability accommodation obligations Employment contracts, common law and collective agreements The Canadian Human Rights Act protects workers in federally regulated employment from discrimination based on grounds that include disability, sex, age, marital status and family status. Provincial and territorial laws provide similar protections, although the exact rules vary. In June 2026, the Canadian Human Rights Commission said organizations should assess human rights risks before using AI in high-impact areas such as employment. It also emphasized that bias testing, transparency and meaningful human oversight should be built into AI systems from the start rather than considered only after harm occurs. An employer does not need to instruct an AI system to discriminate for a discriminatory outcome to occur. A layoff tool may rank workers using projects completed, sales, customer interactions, emails, office attendance, system activity, AI-tool usage or manager ratings. An employee who worked throughout the review period will normally generate more activity than someone who spent part of that period on approved medical or parental leave. Unless the system properly adjusts for the absence, it may incorrectly label the employee as less productive or less engaged. Protected leave can therefore become a hidden negative factor even when the system never directly uses words such as disability, pregnancy or family status. A seemingly neutral rule may create an adverse effect connected to a protected characteristic. Job-protected leave does not make an employee permanently immune from every legitimate restructuring. An employer may sometimes eliminate a position held by an employee on leave when: The restructuring is genuine The position would have been eliminated regardless of the leave The absence was not treated as poor performance The selection criteria were consistently applied Accommodation requirements were considered The employer can document a legitimate business reason Risk increases when leave-related inactivity lowers an employee's ranking or when the timing suggests that leave influenced the decision. Ontario human rights guidance, for example, confirms that disability protections apply to dismissals, layoffs, performance evaluations and return-to-work situations. Employers should confirm the rules applying to their own jurisdiction. 1. Human rights discrimination An automated process may disadvantage employees because of disability, pregnancy, sex, age, family status or another protected ground. Historical data can also reproduce past workplace bias. 2. Failure to accommodate Performance or attendance concerns may be connected to disability. Before relying on those concerns, an employer may need to determine whether accommodation was requested, considered and provided as required by applicable human rights law. 3. Employee privacy AI tools may process health details, leave records, communications, login activity, location information and behavioural predictions. Employers should collect only necessary information, use it for an appropriate purpose, restrict access and explain monitoring practices where required. 4. Unexplainable decisions Employers may have difficulty defending a termination when neither HR nor the manager understands how the system produced its ranking. A confidential vendor formula should not replace a clear business rationale. 5. Rubber-stamp human review A person clicking "approve" does not create meaningful oversight. The reviewer should understand the criteria, verify the information, challenge unusual