Lawyers explain how current Canadian laws around AI in hiring leave ‘a pile of other potential liability issues’ for employers to trip over
More than half of Canadian employers now use AI somewhere in their hiring process — and many believe the technology is making recruitment fairer and more efficient.
However, that confidence may be misplaced. From human rights exposure to untested privacy obligations, the legal risks of AI-assisted interviewing in Canada are real, growing, and largely untested.
In a recent survey of 2,950 job seekers across the US, UK, Ireland, Germany and Australia, 63 percent of respondents had already experienced an AI-led interview. More significantly, 70 per cent said AI was not clearly disclosed before the process began — and 38 per cent had already withdrawn from a hiring process because it included one, found AI-powered platform Greenhouse.
AI job interview disclosure
Ontario's Bill 149 — the Working for Workers Four Act, 2024 — became the first legislation in Canada requiring employers with 25 or more employees to disclose AI use in job postings when it came into force on Jan. 1, 2026.
Matthew McCarthy, employment lawyer at Boughton Law in Vancouver, says the law addresses a real and persistent grievance from candidates.
“The one major complaint from applicants across all industries is not knowing that AI is being used — whether it's at the screening level, at the interview level, wherever,” he says.
Bill 149 is meant to address that base-level issue, but that disclosure on its own resolves very little, says McCarthy.
"It is connected in some ways to the Employment Standards Act, so there should be teeth to it, but all it does is make that base requirement: 'You have to disclose this,'” he says.
“There's still a pile of other potential liability issues that arise."
AI interview tools and data use
Roxanne Davis, managing partner and employment lawyer at Carbert Waite in Calgary, agrees, adding that disclosure from the get-go should be base-line best practice for employers across the country.
Disclosing what tools you’re using and how not only reduces legal risk, she says, but ensures the most qualified candidates make it through the gates.
"You will get the best result, in terms of the person most qualified for the job, if you are setting everybody up on an even playing field... so that they're appropriately preparing based on their understanding of what the process is going to be."
Disclosure obligations already exist under Canadian privacy legislation, which require organizations to explain how personal information — including resumes and recorded video interviews — is collected, used and shared.
In practice, however, Davis says that disclosure can be buried under “generic” privacy statements “that people don't read, and they check a box to move on to a new screen and don't really understand what it is, what the implications are.”
Audit, audit, audit
Interestingly, the Greenhouse report found that candidates perceived nearly identical rates of bias from AI and human interviewers — age bias was flagged by 36 per cent for both, and race or ethnicity bias by 27 per cent for both.
According to Davis, this tracks with what she’s hearing from employers right now.
"The difficulty is that people have a tendency to believe that they are eliminating the bias by using the tools,” she explains.
“People who are using them as screening tools think that they're improving the process because they're eliminating the human bias. Perhaps things won't get screened out, like a lack of an English-sounding name, for example, or credentials from international countries — but because the AI tools are built by humans, they do have those same biases. You just don't see them."
For Canadian employers, that means auditing AI hiring tools isn’t optional, and should be done more often than is probably currently understood; she recommends auditing before deploying any new AI tool and again with every update.
"At this stage, most AI products are being updated every couple of months with new features," Davis says.
"Every time there is one of those updates, it needs to be audited again to see how that filtering process has changed."
Neurodivergence and AI
In the 2026 paper “Screened Out: How Ai Hiring Tools Disadvantage Neurodivergent Candidates”, researchers describe how AI video interview platforms assess candidates on vocal tone, facial expressions, eye contact, and speech cadence — metrics designed according to profiles built from past "successful" employees, who are overwhelmingly neurotypical.
The researchers explain how with autistic candidates, reduced eye contact or flat affect may signal disengagement to the algorithm. For candidates with ADHD, speech disfluencies or tangential answers may also trigger lower scores while not actually measuring qualifications.
A 2025 study, published in the Journal of Law and Society, found the way AI hiring systems are used can create “serious risks of algorithm-facilitated discrimination.”
While this particular bias has not yet been tested in Canadian courts, McCarthy makes it clear that it is only a matter of time.
“If a person has a neurodivergence issue, and it maybe creates an issue interacting with an AI, does that then give rise to a potential discrimination on the basis of disability?" he says.
"I don't know. This hasn't been tested. But do I expect that a claim will be made at some point."
Canadian laws around AI uses
As Canadian HR Reporter has reported, the absence of federal AI legislation since the death of Bill C-27 in January 2025 means human rights codes and PIPEDA remain the primary legal guidelines for in Canada.
For employers that want to stay on the right side of workplace AI regulation in advance of the laws that will inevitably arrive, both Davis and McCarthy recommend transparency, disclosure and regular audits of vendor products as well as their own practices and outputs.
For Davis, the starting point is ensuring AI tools are governed, not just deployed – and by people who understand them.
"It's best to have trained HR professionals do those processes and not let untrained managers do it," she says.
"It's also important to be very thoughtful in the parameters you're giving the AI tool — what you're asking it to screen for, what prompts you're giving it. Look at the before-and-after results to evaluate whether your prompts are achieving what you want to achieve, or whether there is inherent bias built into them that is causing potential human rights risks."
Above all, McCarthy adds, employers underestimate candidate perception as a risk in itself; understanding that, he says, is both a legal risk management strategy and a communication one.
"What are your candidates perceiving, not only your candidates but also your internal people? What do they think about the use of AI?” McCarthy says.
"I think if you get more buy-in, you avoid a lot of the concern. A lot of the concern really is rooted in fear, because so few of us understand what this technology is capable of doing."
