A major US tech company made headlines for firing engineers who refused to adopt AI tools — an approach a Canadian employment law expert does not recommend
The CEO of cryptocurrency exchange platform Coinbase, Brain Armstrong, made global headines recently after revealing he fired engineers who failed to follow his directive to start using artificial intelligence.
However, a Canadian employment law expert cautions that terminating staff for refusing to use AI is far more complex in Canada than in the United States.
“In Canada, the law in terms of terminations is very different from the US,” Michelle McKinnon, a partner with Cassels' Employment Law Group said.
“I do not recommend the Coinbase approach.”
McKinnon notes that in Canada, an employer’s choices are essentially limited to two categories: termination for cause or termination without cause.
“An employer can terminate an employee either for just cause, meaning that you say the employee did something egregious and you’re terminating the employment with immediate effect, or you can terminate an employee without cause, meaning you’re not terminating them for a specific reason – you’re just providing notice that you’re terminating the employment,” she says.
When refusal isn’t insubordination
The question becomes what qualifies as grounds for just cause when AI use is involved.
“If you’ve given lawful and reasonable directions, and you say employees must follow this direction, the employee simply refuses and have no valid justification for that refusal, then depending on the circumstances, you may be looking at a just cause termination,” she says.
But McKinnon warns employers not to leap to that conclusion and says they should ask the employee why they are refusing adoption.
“In 99% of cases, there’s probably going to be a valid or reasonable explanation from the employee,” she explains. “It may be that there’s an underlying human rights issue, a disability of some sort, that is impacting the employee’s ability to adopt the AI or to use the AI tool in its intended purposes.
“AI is new to all of us. I think it’s a bit of an aggressive approach to proceed to an immediate termination, as in the Coinbase case,” she says.
Risks for employers with AI enforcement
One complication arises when employers have no formal policy in place. Without that clarity, a refusal may not meet the legal bar for insubordination. She emphasizes that AI policies must be communicated, acknowledged and paired with training.
“If you are terminating someone for just cause and the employee is not following direction and workplace rules, there has to be clear rules and policies in place that employees understand,” McKinnon explains. “You may have a fabulous AI policy, but if employees are not aware of it, and you haven’t provided training to them on it, they may be able to say they didn’t know.”
The stakes climb higher when AI adoption changes the essence of someone’s job, which could risk a constructive dismissal case, McKinnon says.
“If you’re introducing an AI tool and you say it is now a condition of employment and that new condition is a significant change to how employees currently do their work, then that could result in a constructive dismissal,” she explains.
The line between efficiency and fundamental change is not always obvious.
“If you’re introducing a tool that’s just going to create efficiencies, so employees are still doing the same work, that’s not a constructive dismissal issue,” McKinnon says. “But if you’re introducing a tool and you’re saying, the employee has to use this tool and work in a different way, then I think the employer should assess what is the impact of that new condition.”
For federally regulated workplaces, the risks are even sharper. She points to discontinuation of a function as one example, noting that employers would need to establish that the AI adoption completely eliminated the role.
“Not all employees can be terminated without just cause at the federal level; they can only be terminated for certain specific reasons,” she says. “Terminating an employee is much harder, and it’s more complicated.”
Building an AI policy
Before putting any AI policy in writing, McKinnon urges employers to analyze its true impact and ensure it’s specific to their workplace. She also stresses that confidentiality and intellectual property must be built into policy frameworks.
“If you input confidential information of your business or a client into ChatGPT to generate an outcome, then immediately you are compromising the confidentiality of that information,” McKinnon says.
This goes hand in hand with training, which in her view, is not optional.
“Employees should get good training on these things because it’s not just about using AI to be more efficient. There are risks associated with using these tools,” she says.
That training also needs to address what happens when things go wrong.
“If an employee was approved to use an AI tool, if they realize that the company’s confidential nature has been compromised, then there needs to be a process set out in the policy for them to report that breach as quickly as possible,” she explains.
Looking ahead, McKinnon sees the same flashpoint repeating itself.
“The key potential claim is constructive dismissal,” she says. “Employees may say, I’m refusing to do my work in this way because it fundamentally alters the nature of my job, and so I’m going to resign and claim constructive dismissal.”
