Ontario’s AI Hiring Disclosure Is Now in Force: What Employers Need to Review

Workplace AI Canada — Ontario’s AI hiring disclosure is now one part of a wider governance review: employers should map tools, document human oversight, protect personal information, and check provincial obligations before relying on automated recruitment.

Leah Mercer

Leah Mercer

— 4 min read
Share
Editorial image for Workplace AI Canada
Photo by Taylor Vick on Unsplash.

What changed on January 1, 2026

Ontario’s new public job-posting requirements came into effect on January 1, 2026. The rules amend the Employment Standards Act, 2000 and include disclosure requirements for certain employers using artificial intelligence in recruitment.

The reported requirements apply to employers with 25 or more employees on the day a posting is made. A publicly advertised job posting is an external posting advertised to the general public in any manner. Where covered, the posting must disclose whether artificial intelligence is used to screen, assess, or select applicants.

The same package includes other requirements, including pay-transparency and vacancy-status disclosures, notification of hiring decisions within 45 days, and retention of job postings and applications for three years. The materials reviewed state that the AI disclosure requirement does not cover general recruitment activity that does not advertise a specific position, internal job postings, or postings for roles performed outside Ontario.

The available commentary does not identify detailed government guidance on the wording or format of the AI disclosure. It also describes this particular rule as requiring no bias audit, no explanation of the system’s operation, and no candidate notice at the decision stage. Those statements describe the reported scope of this rule, not a conclusion that other legal or organizational responsibilities do not apply.

Why this matters beyond the posting

Canada’s workplace AI environment remains a patchwork rather than one uniform framework. Existing privacy, employment, human-rights, intellectual-property, consumer-protection, and tort laws may apply to AI-related risks. Regulators may also issue binding or non-binding guidance, while provinces may create targeted duties through existing statutes.

For employers, the Ontario posting statement is therefore only one control point. Human-rights obligations may apply when an AI system produces an unjustified adverse differential impact on a prohibited ground. The risk may be greater when a system is used systematically or affects many employees. The available research states that employers may remain accountable for discriminatory effects during onboarding, employment, or termination, even where a technology supplied or influenced the decision.

Privacy obligations may also arise when technology uses personal information to make or support automated decisions about employees. The research describes Quebec’s automated-decision rule as applying where a decision is based exclusively on automated processing. Meaningful human involvement may take a decision outside that particular provision, while other profiling, privacy-by-default, transparency, necessity, or proportionality requirements may still be relevant.

The research also reports Quebec enforcement guidance issued in 2025 concerning automated decision-making and algorithmic-management tools in employment. It states that employers deploying AI systems affecting employees’ rights may face investigation or penalties where transparency, necessity, and proportionality expectations are not met. This is an enforcement signal, not a prediction about any particular employer.

At the federal level, the materials identify a Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems and a 2025 Implementation Guide for Managers of Artificial Intelligence Systems. The available context describes these as federal instruments that do not generally bind private employers. The broader research also reports a late-2025 federal privacy investigation involving algorithmic recommendations to children, illustrating that algorithmic design choices can attract accountability and fairness scrutiny under privacy law.

A practical review for employers and HR teams

Use this as a general governance review, not individualized legal, employment, privacy, or compliance advice.

  • Inventory every recruitment tool that screens, assesses, ranks, recommends, or selects applicants. Record the vendor, purpose, data used, decision point, human role, and jurisdictions involved.
  • For each Ontario posting, record the employer headcount on the day the posting goes live. Confirm whether the posting is external, publicly advertised, for a specific position, and connected to work performed in Ontario.
  • Update controlled posting templates so the AI-use disclosure, pay-transparency statement, and vacancy-status statement are considered together where applicable. Do not assume that a generic statement accurately describes the tool’s actual use.
  • Preserve the posting, application materials, approval record, and relevant system or vendor version information according to the applicable retention process. The research identifies a three-year retention requirement for postings and applications under the reported Ontario rules.
  • Confirm that the process for communicating hiring decisions can meet the reported 45-day timeline. Assign ownership rather than leaving the task solely with a recruiter or an automated workflow.
  • Document where a person reviews, overrides, or validates an AI-supported recommendation. Human involvement should be genuine and recorded; it should not be treated as a substitute for reviewing applicable privacy, human-rights, or employment obligations.
  • Check whether applicant or employee personal information is being used for automated decision-making, profiling, or model improvement. Map the data flows, notices, access controls, vendor terms, and deletion or retention practices.
  • For Quebec operations, obtain jurisdiction-specific review of transparency, necessity, proportionality, profiling, and automated-processing requirements. For other provinces and territories, do not infer that the absence of an AI-specific statute means the absence of existing-law obligations.
  • Establish an owner for monitoring statutory changes, regulator guidance, vendor changes, complaints, accommodation needs, and potential discriminatory outcomes. Accessibility and human-rights review should be part of deployment governance, not an afterthought.

Limits, uncertainty, and evidence gaps

The Ontario disclosure requirement is described consistently in the independent legal commentary supplied for this issue, but the research context does not include a direct official Ontario statute or regulation link. Official Canadian sources are required for laws and regulations under this publication’s source policy. Readers should therefore verify the current Employment Standards Act, 2000, associated regulations, and government guidance before relying on operational details.

The available sources also disagree or use different levels of certainty about the status and scope of some provincial instruments. One legal guide uses cautious language such as “may require” for the Ontario disclosure obligation, while independent summaries describe the requirement as in force. The scope of Quebec’s automated-decision provisions, the implications of meaningful human involvement, the outcome of the federal privacy investigation, and future provincial developments remain subject to change.

This issue does not establish whether a particular tool is lawful, fair, accessible, necessary, proportionate, or appropriate for a particular workplace. It does not replace qualified legal, privacy, human-rights, accessibility, or technical advice. Employers should evaluate their own facts, agreements, policies, and jurisdictional obligations.

Sources

Research limitation: no direct official Canadian primary-source URL for the Ontario statute or regulation was included in the supplied context.

Read more