As AI-powered safety monitoring tools spread into Canadian workplaces, experts warn HR leaders to tread carefully
A growing number of Canadian employers are deploying artificial intelligence to monitor workplace safety – tracking truck driver fatigue, flagging ergonomic risks, and scanning job sites for hazards. But the same data streams that promise to protect workers can just as easily become instruments of control, and HR leaders are under mounting pressure to draw a clear line between the two.
A recent report by the US Government Accountability Office found that digital surveillance of employees improved safety, but also increased injury risk and affect mental health if productivity is pushed. Yet AI adoption continues to accelerate.
AI surveillance is already a legal minefield, raising complex questions around privacy rights, constructive dismissal, and worker dignity.
Transparency as the non-negotiable baseline
For Wyatt Tessari L’Allié, Founder and Executive Director of AI Governance and Safety Canada (AIGS Canada), the most critical factor on the efficacy of AI-driven safety tools isn’t the technology itself – it’s how it’s introduced.
“The worst thing to do is have a top-down approach and impose this on your employees and make it feel like you’re surveilling them, even if it’s for good purposes,” says L’Allié. “There’s already a lot of backlash broadly against AI, and particularly against AI as a C-suite project being pushed down on everyone else – the most important thing is to make sure that if you’re going to put in place those kinds of systems, they’re done in collaboration with people, not against them.”
L’Allié also urges HR leaders to interrogate the technology before deploying it. “It’s important to be very mindful of what exactly you’re trying to track, what outcomes you’re trying to achieve, and do you actually need technology for that – and if so, which technology and why?” he says.
Mike Russo, a registered occupational hygienist and owner of WorkBright, has optimism for AI-powered safety tools – with caveats. “I think there is tremendous opportunity in reducing risks by actively monitoring hazards,” says Russo. “But many HR leaders have to acknowledge what the concerns are, primarily privacy – that’s going to be on the minds of many employees who are part of that AI world and where it's being used to actively monitor hazards in the workplace.”
If a tool is deployed for safety, that’s what it must be used for, full stop, says Russo. “If a company says AI will be used for actively monitoring hazards, then that’s all it should really be used for,” he says. “Where trust erodes is where you start to see monitoring technologies being used for other purposes.”
The data dual-use problem
Dilara Baysal, a Postdoctoral Researcher at Bishop’s University who studies the effects of productivity and AI-driven technological change on labour institutions, pinpoints another growing risk: the temptation to repurpose safety data for performance management.
“Employers are often really tempted to use footage for productivity, for performance, for disciplinary action,” says Baysal. “It’s very important that footage is used only and exclusively for safety purposes, and that it’s written and communicated to employees – that should be the line when it comes to AI monitoring.”
Dilara refers to the 2025 Quebec arbitration case involving transportation company Coach Canada as an example: the company used Samsara technology to record drivers without disclosing to workers or their union that it was also extensively tracking biometric data – footage later widely circulated internally. “This is how not to do it, basically,” Baysal says.
She’s equally skeptical of software promising productivity gains from employee data. “I would really warn against products that promise to use your employees’ data in the name of productivity,” she says. “Resist the urge to think you’re sitting on a gold mine of data. That’s a very, very dangerous place to be.”
Operating in a legal grey zone
Ontario requires employers with 25 or more employees to have a written electronic monitoring policy, and Quebec has advanced its own privacy protections. But at the federal level, there’s still no legislation governing how AI-generated data can be used against workers in disciplinary or dismissal proceedings.
“It does leave companies in a bit of a void,” L’Allié says. “Transparency and good data protection practices – being clear and open about what you’re doing – is very important, so that if you’re making a mistake, it can be caught early.”
Baysal contrasts this with the EU, where Italy, Spain, and Germany require worker consent – and in some cases co-determination rights – before certain technologies can be introduced. In Italy, some biometric surveillance remains illegal even with worker consent, recognizing that employees are rarely truly free to say no, she says.
“In Canada, there is a lot of responsibility on HR leaders around what’s reasonable and what’s not,” says Baysal. “If you are being monitored all the time, how autonomous are you going to feel at the workplace? How much confidence and trust are you going to have, and how much motivation will you have to build expertise, or are you going to be too concerned about not getting caught doing something?”
She also flags the equity dimension. Unions have warned that lower-income workers, younger employees, people with disabilities, and racialized workers face disproportionate AI monitoring – and Baysal cautions that technology cannot substitute for fixing underlying culture. “AI tools are notoriously famous for discriminating against people of colour and people with disabilities,” she says. “If you think AI is going to solve all your problems when it comes to people management, it’s already not a good place to start.”
Don’t forget the human element of AI
“There is a very important human element to managing health and safety,” says Russo. “That human connection is very important in translating information to employees and motivating them to want to adopt certain behaviours or control measures. It’s something that shouldn’t be forgotten when working with AI in any capacity.”
L’Allié points to a dimension HR leaders need to keep in their purview: autonomous AI agents that themselves require oversight. “Increasingly, what you need to monitor is not just the employees, but also the AI agents that employees are using,” he says. “Don’t forget to monitor the technology that you’re using to monitor others.”
Insufficient guardrails on the use of AI tools will aggravate existing issues, according to L’Allié. “Companies that have low trust and already have high turnover will use these tools even more – and cause even more turnover,” he says. “The technology is not so much the root cause, but the symptom of the overall challenges within the organization.”
Baysal suggests organizations should have workers involved in the implementation of AI safety tools from the beginning. “It’s very important that this kind of software isn’t taught only by the HR departments or management, but with the involvement of worker representatives who can also point out ways that the software can fail or give different ideas on like how to integrate it into the process.,” she says.
L’Allié recognizes the potential with AI-powered safety tools and monitoring but cautions that it’s a rapidly evolving technology. “There are no perfect solutions – everything with this technology is going to need to be iterated over time, he says. “Don’t aim for perfect – but do aim for full transparency.”
