AI in Hiring: Disclosure and Discrimination Risks
Employers increasingly use AI in hiring to source, screen, rank and interview candidates. Canadian law does not ban these tools, but it imposes disclosure duties in Ontario and Quebec, prohibits discriminatory outcomes everywhere, and requires privacy compliance. Employers remain responsible for what their vendors' tools do.

Key takeaways
- Ontario requires employers with 25+ employees to disclose AI used to screen, assess or select applicants in public job postings from 2026.
- Quebec requires notice when a decision is based exclusively on automated processing of personal information.
- Human rights codes prohibit discriminatory hiring outcomes, including those produced by AI tools.
- Privacy laws require transparency and, in many cases, consent for collecting candidate data.
- No federal AI statute is in force; Bill C-27's AI act died in January 2025.
- Human review, bias testing and vendor due diligence are the core safeguards.
How employers use AI in hiring
AI in hiring takes many forms: tools that write job descriptions, source candidates from online profiles, parse and rank resumes, score video interviews, administer gamified assessments, run chatbots that pre-screen applicants, and predict which candidates will succeed or stay. Many applicant tracking systems now include AI ranking features by default. These tools promise efficiency and consistency but can reproduce or amplify bias in historical data, and they often operate in ways that employers do not fully understand.

Canada has no comprehensive AI statute in force. The proposed Artificial Intelligence and Data Act, part of Bill C-27, died when Parliament was prorogued on 6 January 2025. Its successor privacy bill, Bill C-36, introduced on 15 June 2026, does not include a standalone AI act and had not received royal assent. In the meantime, AI in hiring is governed by employment standards in Ontario, private-sector privacy laws, human rights codes and general employment law.
Ontario's AI disclosure rule
From 1 January 2026, s. 8.4 of Ontario's Employment Standards Act, 2000 (ESA) requires employers with 25 or more employees that advertise a publicly advertised job posting and use artificial intelligence to screen, assess or select applicants for the position to include a statement disclosing that use. The ministry's guidance explains that the regulation defines AI broadly as a machine-based system that infers from inputs how to generate outputs such as predictions, content, recommendations or decisions. The disclosure duty applies whether the employer built the tool or bought it from a vendor.
| Jurisdiction | Rule on AI in hiring | Status |
|---|---|---|
| Ontario | Disclose AI used to screen, assess or select applicants in public postings (25+ employees) | In force 1 January 2026 |
| Quebec | Notice where a decision is based exclusively on automated processing; right to submit observations | In force under Law 25 |
| Federal | No AI statute in force; PIPEDA applies to federally regulated employers' candidate data | Bill C-27 died January 2025; Bill C-36 pending |
| All provinces | Human rights codes prohibit discrimination in hiring | In force |
Disclosure is a floor. It does not make a discriminatory tool lawful, and it does not satisfy privacy notice requirements. It does, however, put candidates on notice, which may increase questions and complaints about how decisions were made.
Quebec's automated decision rules
Quebec's private-sector privacy law, as amended by Law 25, requires an organization that uses personal information to render a decision based exclusively on automated processing to inform the person of that fact, no later than the time it informs them of the decision. On request, it must tell the person what personal information was used, the reasons and principal factors and parameters that led to the decision, and the right to have the information corrected.
The person must be allowed to submit observations to a member of staff who can review the decision. Where AI screening automatically rejects Quebec applicants without human involvement, these rules apply. Keeping a human in the loop changes the analysis, but only if the human review is real.
Human rights and discrimination
Every Canadian jurisdiction prohibits discrimination in employment, including in recruitment and hiring, on grounds such as race, sex, age, disability, religion and family status. Discrimination does not require intent; a neutral practice that has an adverse effect on a protected group is discriminatory unless justified as a bona fide occupational requirement. AI tools trained on historical hiring data can penalize gaps in employment that correlate with parental leave or disability, favour language patterns associated with particular groups, or rate video interviews differently by accent or appearance. An employer that uses such a tool is responsible for the result, even if a vendor designed it.
Accessibility is part of the analysis. Timed online assessments, video interviews and gamified tests can disadvantage candidates with disabilities unless accommodations are offered. Employers should tell candidates how to request accommodation and ensure that alternatives are available. Ontario's Human Rights Code and the Accessibility for Ontarians with Disabilities Act standards on employment also apply.
Ask vendors for evidence of bias testing on populations comparable to your applicant pool, the features the model uses, and how often it is retrained. Run your own adverse impact analysis on outcomes by stage of the process. If a tool cannot be explained or tested, do not let it make or effectively make rejection decisions.
Privacy obligations
Candidate data is personal information. Federally regulated employers are subject to PIPEDA for that data; Alberta, British Columbia and Quebec have private-sector privacy statutes that apply to provincially regulated employers. Transparency is common to all: candidates should be told what information is collected, from which sources, for what purposes, and whether automated tools are used. Scraping social media or using third-party data brokers to build candidate profiles raises consent and accuracy issues. Video interview analysis can involve biometric information, which privacy regulators treat as highly sensitive.
Data minimization and retention limits matter. Keep candidate data only as long as needed for the competition and any legal requirements, and do not reuse it to train vendor models without a clear legal basis. Cross-border data transfers to vendors' servers abroad require transparency and, in Quebec, a privacy impact assessment before personal information is communicated outside the province.
Vendor due diligence and contracts
- Understand what the tool does at each stage and whether it ranks, filters or rejects candidates.
- Obtain documentation of training data, validation, bias testing and updates.
- Require compliance with Canadian privacy laws, data location commitments and breach notification.
- Prohibit use of candidate data to train the vendor's general models without consent.
- Secure audit rights and access to outcome data for your own adverse impact testing.
- Allocate responsibility through warranties and indemnities, recognizing that regulators and tribunals will still look to the employer.
Human review and documentation
Meaningful human oversight is the most important control. Recruiters should review AI recommendations, understand the factors behind them and have authority to depart from them. Rejections at early stages should be sampled and reviewed for patterns. Decisions should be documented so that, if a candidate complains to a human rights tribunal, the employer can explain why the candidate was not selected and show that the explanation does not rest on a protected ground. Our insight on human rights tribunal claims discusses defences.
Generative AI in recruitment workflows
Generative AI is now common in recruitment even where no formal screening tool is used. Recruiters draft job postings with chatbots, summarize resumes, generate interview questions and write rejection letters. These uses carry their own risks. Job descriptions generated by AI can include requirements that are not genuinely needed, such as unnecessary degree or experience requirements, which may have an adverse effect on protected groups or, in Ontario, may amount to a prohibited Canadian experience requirement. Pasting candidate resumes into public AI tools can disclose personal information to third parties in breach of privacy law and the employer's own policies.
Whether generative tools used by recruiters count as AI used to "screen, assess or select" applicants under the Ontario rule depends on how they are used. Summarizing a resume for a recruiter who then decides is different from asking a tool to rank candidates. Employers should set internal rules on approved tools and uses, and decide on a consistent approach to disclosure.
What candidates can do
Candidates who believe AI in hiring treated them unfairly can file human rights applications, privacy complaints with the relevant commissioner, and, in Ontario, complaints about non-compliance with posting rules. In Quebec, they can ask for the factors behind an automated decision and submit observations. Collective complaints are possible where a tool affects many applicants in the same way. Employers that can produce clear documentation of how a decision was made, and evidence that the tool was tested and supervised, are far better placed to respond.
AI beyond recruitment
The same issues arise when AI is used to allocate shifts, assess performance, recommend promotions or identify employees for layoffs. Selection for termination using a model that disadvantages older workers or employees on leave can create human rights liability on top of wrongful dismissal claims. Monitoring tools with AI analytics must be described in Ontario's electronic monitoring policy where applicable, as explained in our insight on electronic monitoring.
A governance framework for AI in hiring
- Inventory every AI-enabled tool used in recruitment and employment decisions.
- Classify tools by impact: those that reject or rank candidates need the most controls.
- Update job posting templates for Ontario disclosure and privacy notices for all jurisdictions.
- Implement Quebec automated decision notices and review processes where applicable.
- Test outcomes for adverse impact and offer accommodation routes.
- Assign accountability to a named executive and review the program annually.
- Train recruiters and hiring managers on the tools' limits, the disclosure rules and how to handle accommodation requests.
- Keep records of postings, notices, reviews and testing results so the program can be explained to a regulator or tribunal.
Illustration
A national logistics company with 600 employees uses an applicant tracking system whose AI feature ranks applicants for warehouse supervisor roles and automatically declines the bottom half. For Ontario postings, it must disclose that use. For Quebec applicants who are declined automatically, it must give notice that the decision was based exclusively on automated processing and offer a review.
An internal audit shows that applicants with employment gaps of more than a year are ranked lower, which disproportionately affects women returning from parental leave. The company switches off automatic rejection, adjusts the model's inputs with the vendor, has recruiters review every ranking, and documents the changes. Those steps reduce the risk of human rights complaints and show regulators that AI in hiring is being supervised.
The example also shows why outcome testing matters. The bias was not visible in the vendor's marketing materials and appeared only when the employer compared outcomes for different applicant groups.
Federal and international outlook
Although no federal AI act is in force, regulators are active. Privacy commissioners have issued principles for responsible generative AI, and human rights commissions have published guidance on algorithmic discrimination. Multinational employers may also be subject to foreign rules, such as the European Union's AI Act, which treats AI used in recruitment as high-risk, and US state and local laws on automated employment decision tools. A single global governance framework that meets the strictest applicable standard is often the most efficient approach.
How TRW & Co helps
We advise employers and HR technology providers on AI in hiring: disclosure and notice requirements, privacy compliance, bias testing protocols, vendor contracts and defence of complaints. Our employment law, artificial intelligence and privacy teams work together. See also our insight on Ontario employment standards changes. This page is general information, not legal advice. Contact us to review your hiring tools.
Frequently asked questions
Do Ontario employers have to disclose AI use in hiring?
Yes, if they have 25 or more employees and use artificial intelligence to screen, assess or select applicants for a publicly advertised position. From 1 January 2026, the posting must include a statement disclosing that use. The definition of AI is broad.
What does Quebec require for automated hiring decisions?
Where a decision is based exclusively on automated processing of personal information, the employer must inform the candidate, explain on request the information used and the principal factors, and allow the candidate to submit observations to a staff member who can review the decision.
Is there a federal AI law?
Not in force. The Artificial Intelligence and Data Act in Bill C-27 died on prorogation on 6 January 2025. Bill C-36, a privacy bill introduced on 15 June 2026, does not contain a standalone AI act and had not received royal assent.
Can an employer be liable for a vendor's biased tool?
Yes. Human rights law focuses on discriminatory effects, not intent, and employers are responsible for their hiring decisions, including those made with vendor tools. Contracts can allocate costs, but they do not shift legal responsibility to the vendor.
What safeguards should employers adopt?
Meaningful human review of AI recommendations, adverse impact testing by stage, accommodation routes for candidates with disabilities, transparent privacy notices, vendor due diligence with audit rights, and clear accountability for the program.
Sources
General information only, not legal advice. References to statutes, courts and authorities describe published law and practice at the date shown.


