Artificial intelligence is no longer a future-state issue: it's operating inside Canadian businesses already, and the legal and operational landscape surrounding AI in the workplace is moving faster than most governance frameworks can keep pace with.
On June 1, 2026, BLG's Labour and Employment lawyers gathered coast to coast to work through four areas that demand immediate attention:
- The evolving legislative landscape around AI;
- The risks of uncontrolled early adoption;
- The human-rights exposure embedded in AI-driven hiring; and
- The intersection of AI with performance management.
1. The legislative landscape
Ryan Mulders, Associate, Ottawa
Canada's legislative framework for regulating AI in the workplace is still developing, and an area that is expected to grow in the near future at the both the federal and provincial level. To date, Ontario is the only jurisdiction in Canada that has legislated on AI issues within its employment standards legislation. However, employers may also have certain obligations by virtue of existing privacy legislation, and may be subject to new requirements as a result of proposed and forthcoming legislation.
What's in force today
In the privacy context, Québec’s Private Sector Act regulates “automated processing” of personal information and requires private sector employers to disclose personal information that is being used to render a decision. Similarly, Alberta’s Protection of Privacy Act requires public-sector employers to disclose if they intend to input personal information into an “automated system” for the purposes of making certain decisions. In both cases, the legislation would likely apply to AI systems that are used to make hiring, firing and performance management decisions where personal information is being used.
In the employment context, as of Jan. 1, 2026, Ontario's Employment Standards Act, 2000 (ESA) now provides that a “publicly advertised job posting” will need to disclose the use of AI if the employer uses AI to “screen, assess or select applicants for the position.” The ESA also provides that employers with 25 or more employees must maintain a written electronic monitoring policy, and generally requires those employers to account for any AI tools used to monitor employees in the workplace within the policy.
Recent developments
At the provincial level, Nova Scotia's Bill 234 proposes amendments to employment standards legislation that would require employers to disclose the use of AI in job postings in a manner similar to Ontario, and Manitoba recently passed Bill 51, the Public Sector Artificial Intelligence and Cybersecurity Governance Act, regulating the use of AI by public sector entities. Additionally, the federal government recently introduced its new AI for All strategy and is expected to address AI impacts on the labour market.
Internationally, the European Union’s Artificial Intelligence Act may apply to employers located in Canada and elsewhere who are recruiting within the EU, and may require compliance with certain assessment, mitigation, oversight and security requirements as early as August 2026.
Key takeaways
Employers in Ontario should assess if and how AI is being used to make decisions about their employees, and address any gaps in required policies that may result from the recent adoption of new AI technologies. Although this area will continue to develop, the current landscape also suggests that AI issues in the workplace may be addressed though transparency and disclosure obligations (although oversight and other requirements may also develop), so employers should also be prepared for these types of changes in the future.
2. Challenges to early AI adoption: Get visibility before the risks get ahead of you
Jessica Wuergler, Senior Associate, Toronto
As of 2026, the question is no longer whether employees are using AI in your organization. They are. The urgent question is whether you know what tools are being used, what information is being entered, and who is accountable for the output.
Shadow AI is a widespread problem. Without an AI policy, employees adopt free, convenient tools without oversight, creating inconsistent practices, uncontrolled data-sharing, and an evidentiary record the organization doesn't control. When that surfaces in litigation or a regulatory proceeding, the exposure is yours.
AI models don't think, they predict. Generative AI produces polished, confident text. That doesn't make it accurate. The model generates probable responses based on patterns, not professional or legal judgment. An AI tool might omit a key allegation from a complaint summary, overstate performance issues in a discipline letter, or produce a policy summary that doesn't reflect your actual policy. Unreviewed AI output cannot form the basis for a defensible workplace decision.
Risk mitigation requires two gates. On the front end, assess and vet tools before authorizing them, including vendor assurances on data retention and storage. On the back end, require human review of all AI outputs. A good tool used badly is still a problem.
An AI policy is not optional. To be clear: if you lack an AI-specific workplace policy, you are already operating with unmanaged – and indefensible – risk. A defensible policy defines approved tools; sets rules for what information may be entered; requires human oversight; establishes reporting mechanisms; and sets out enforcement consequences. The goal is simple: tell employees exactly what to do, and what not to do, before they improvise.
Key takeaway
Both front-end tool assessment and back-end user accountability are required. Without both, governance exists on paper only.
3. AI and human rights: The recruitment stage is the first risk zone
Lisa Carlson, Senior Associate, Vancouver, and Tommy Leung, Senior Associate, Calgary
AI is widely used in recruitment: drafting job descriptions, screening applications, managing candidate communications and more. But where AI is making or meaningfully influencing those decisions, employers may be inadvertently violating human rights legislation without knowing it.
Every Canadian province protects job applicants and candidates against discrimination on specified grounds. While Canadian tribunals have not yet published decisions on AI-driven recruitment discrimination, U.S. litigation offers a clear warning of what's to come:
- EEOC v. iTutorGroup: Software automatically rejected female applicants aged 55 or older and male applicants aged 60 or older. The case settled for US$365,000 paid to over 200 affected candidates.
- Mobley v. Workday, Inc.: An ongoing class action alleging Workday's AI screening platform discriminated on the basis of race, age and disability. The court rejected Workday's argument that it was merely a software vendor, finding it acted as an agent of its employer clients.
- Elegant Enterprise-Wide Solutions: An AI-generated job posting included citizenship status restrictions. The Department of Justice (DOJ) found this amounted to immigration-status discrimination and imposed three years of monitoring requirements.
Key takeaway
Conduct bias testing before and during deployment of any AI recruitment tool, maintain human oversight at the final hiring stage, and audit AI-generated job postings before they go live.
4. Performance management and AI: Managers must remain the decision-makers
Kabrina Péron, Associate, Montréal
The fundamentals of performance management remain unchanged by AI. Employees must be informed of expectations and shortcomings, given adequate support and sufficient time to improve, and warned of the risk of termination if no improvement occurs. These obligations exist regardless of the technology involved.
In this, AI can play a useful supporting role with analyzing performance data, summarizing evidence, generating feedback, and reducing administrative workload. However, the risks are real and specific:
- Proxy data (attendance, email frequency, time online) can systematically disadvantage employees on protected leave or managing a disability, without any discriminatory intent.
- AI misses higher-order contributions such as mentorship, leadership, or conflict resolution that don't appear in datasets.
- Polished language is not reliable evidence. AI-generated disciplinary letters converge toward a confident, assertive tone. When the underlying evidence fails to support that severity, the gap creates a significant liability.
- AI-generated disciplinary recommendations will not account for progressive discipline principles, collective bargaining obligations, employee tenure, or the risk of complaints before labour- and human-rights bodies.
A developing concern from U.S. proceedings: courts are ordering disclosure of the AI prompts used to generate workplace assessments. Those prompts are traceable and potentially admissible. Managers should be thoughtful about how they frame instructions to AI tools.
Key takeaway
AI is a starting point, not a decision-maker. Managers must retain full interpretive judgment and be able to justify every decision, independently of what any AI tool recommended.
AI and HR in Canada: Short answers to frequent questions
- How to handle AI-enabled recording devices (such as smart glasses) in the workplace? Disclose in electronic monitoring policies (Ontario) and privacy policies. Avoid always-on recording and use clear warning indicators when recording is occurring. For more information, see our January 2026 Insight: Smart glasses at work: Heads-up innovation, head-on compliance for Canadian HR & legal.
- When does an AI-driven job change trigger constructive dismissal? When AI implementation unilaterally changes fundamental employment terms like duties, compensation or hours, constructive dismissal risk arises. Obtain written consent with fresh consideration or provide advance written notice before implementing material changes.
- Can AI fluency be required of existing employees? Yes, but carefully. If it wasn't part of the original role, imposing it without proper process may constitute a fundamental change to employment terms. Provide training, connect the requirement to actual job duties, apply it consistently, and allow adequate time to adapt.
- Should AI note-taking be disclosed in meetings? Check your policy first. As a best practice, disclose AI-scribe use at the outset of any meeting. Consider whether transcription is actually appropriate: in sensitive conversations, its presence may cause employees to disengage or withhold information entirely.
How BLG can help
Navigating the gap between the rapid adoption of AI in the workplace and the slower pace of legislation is the primary challenge facing Canadian employers today. BLG’s Labour and Employment Group can help you move from risk to readiness across all four areas covered in this article:
- Policy and governance: Building the foundational AI policy your organization currently lacks;
- Hiring and recruitment: Assessing your tools and processes for compliance with new disclosure and human rights obligations;
- Performance management: Equipping your managers to use AI as an effective tool, not an unaccountable decision-maker; and
- Constructive dismissal and change management: Implementing AI-driven changes to job roles with a clear and defensible process.
