The government of Ontario has taken a sensible first step in requiring publicly advertised job postings to include any role artificial intelligence (AI) may have in hiring. Employers with 25 or more employees and who use artificial intelligence to screen, assess or select applicants must say so. That’s good.
But this reform also shows how narrow the public conversation remains. It makes AI visible at the point of hiring, but leaves too much of workplace AI in the dark as to how it assigns work, measures performance, sets targets and triggers discipline. Canada’s recently announced artificial intelligence strategy doesn’t go any further. It treats AI as a democratic concern and names hiring as one area where AI systems will make “consequential decisions about Canadians’ lives.”
Yet it frames workplace AI mainly around skills, adoption and productivity. That is where the deeper problem lies. AI does not just help manage work. It writes the rules.
Public anxiety about AI and work usually settles on one pervasive fear: Machines will wipe out jobs and leave workers behind. But technology does not simply eliminate work. It reorganizes it. Some jobs disappear. New ones emerge. Entire sectors are remade. But the deeper shift is harder to see and easier to normalize. Decisions once made by supervisors, managers and human resources staff are increasingly shaped by computers that sort, score, predict and nudge. The real threat is not unemployment. It is the steady erosion of workers’ power, autonomy and bargaining capacity.
Surveillance is no longer a big enough word for this problem. Surveillance implies distance. Work is watched from the outside, recorded after the fact and then acted on. AI-driven management goes further. It enters the workplace’s decision-making. Worker data is fed back into scheduling, task assignment, performance evaluation and discipline. AI machinery does not simply observe work relations from the outside. It helps generate those relations from within the very systems that assign, judge and discipline work. What begins as a record becomes a rule.
Workplace data has a second life
This is the logic of surplus data. Workplace data once described work after the fact. A keystroke, a location ping, a customer rating or the time taken to complete a task recorded something that had already happened. Under AI-driven management, that record is no longer exhausted by its first use. Its surplus lies in its second life. Data is accumulated, recombined and returned to the workplace as a score, ranking, target or threshold. Data no longer simply records what workers have done. It decides what happens next.
Data traces become productivity thresholds, scheduling priorities, promotion filters and disciplinary flags. Workers who slow down because the pace is unreasonable may be classified as less productive. That classification can be used to justify fewer shifts, closer monitoring or discipline. Workers adapt. That adaptation creates more data. The loop continues.
That is what is new. Employers have always measured performance. But measurement used to be more closely tied to human judgment about what a worker did, why it happened and what context mattered. Under AI-driven management, measurement can become a moving rule. Thresholds can be raised, classifications revised and standards changed without workers knowing how or why. Sometimes this happens without explanation. Sometimes it happens without notice. Over time, these tools do not simply enforce workplace standards. They become part of the machinery that creates the standards by which workers are governed.
Algorithms are becoming workplace rules
This is not speculative. Evidence before policymakers already shows how AI-driven management is changing work. In a recent submission to a Senate committee, the Canadian Union of Public Employees (CUPE) warned that AI-amplified monitoring and algorithmic management are shifting workplace power toward employers. The submission shows how workplace data can be processed into profiles, predictions and decisions about hiring, task assignment, promotion, wage-setting, discipline and termination. It also warns that these tools can intensify work, increase anxiety and harm workers’ psychological well-being.
Policymakers and regulators must pay more attention to this “surplus” in data. The federal government says AI should reflect Canadian values, including the protection of human rights and democracy. That commitment already belongs in the national conversation about AI. But it must be brought more directly into the workplace, where AI now shapes income, dignity, privacy and bargaining power most directly. Yet this remains one of the weakest parts of the public debate.
A registry would make hidden rules visible
That is why Canada needs more than hiring disclosure. It needs a registry of labour algorithms. Such a registry would make hidden workplace rules visible, accountable and open to challenge. It would not require government to seize source code, expose legitimate trade secrets or pretend Ottawa can regulate every workplace. The duty would fall on employers that use AI or algorithmic tools, even if the technology were supplied by a multinational vendor.
Any employer using the above to decide, rank, score, recommend or significantly shape hiring, scheduling, pay, productivity targets, discipline or dismissal would have to file a standardized record with a government regulator. That record would need to include what the system did, what worker data it used and which decisions it affected. It would also have to explain whether the employer’s rules, thresholds or data sources had changed over time, who inside the organization was responsible for human review and how workers could challenge an outcome. Filings would have to be updated when systems materially changed.
Canada’s labour protections aren’t ready for the age of AI
Ottawa could begin by using federal contracts and labour reporting systems to set the standard. Provinces and territories could adopt the model through their own labour, privacy and employment laws. Public entries in the registry would allow workers, unions and job applicants see how AI and labour algorithms are shaping their working lives. Confidential filings would give regulators enough detail to investigate harm.
The goal is not to ban workplace technology. It is to ensure that workers are not governed by AI-shaped rules they cannot see, question or contest. A registry would not solve everything, but it would force AI-driven hidden rules into the open before hardening into the everyday terms around pay, scheduling, discipline and dismissal.
The central labour question in the AI era is not whether machines will replace us. It is whether workers will retain meaningful power over the rules that shape their lives.
