On September 25, 2026, the University of Kentucky approved an AI exemption at work for research administrator David Durman, excusing him from personally using AI in his job on the basis of his humanist beliefs after its Office of Equal Opportunity found no undue burden on his department, according to the American Humanist Association’s October 26 release. His employer had not yet required anyone to use AI. Our view is that Canadian employers should pay attention for a reason that has little to do with faith: the wording of an AI expectation decides whether a request like this ever reaches a tribunal, and an expectation written as an outcome rather than a usage count prevents most of the problem from forming.
What Kentucky’s AI exemption at work covers, and what it does not
Kentucky approved one employee’s request to do his job without personally using AI, after its equal opportunity office found no undue burden. It did not decide that humanism always qualifies, it set no rule for other staff, and by its own account to reporters it kept the right to revisit the arrangement as the technology and the job change.
The sequence in the association’s release is short: a written request in late August, a meeting in mid-September and an approval on September 25. The association’s staff attorney says humanism is entitled to the same protection under Title VII as traditional religions, which is the view of the organization that backed the request; no court or agency has ruled on it. A university spokesperson told TechCrunch on October 29 that accommodations “may be reviewed to ensure they remain reasonable and workable,” and that colleagues and leadership had recommended the tools without mandating them.
Kentucky follows at least one earlier case this year. In June, Futurism reported that a Unitarian Universalist software engineer at a North Carolina company had an exemption from AI use approved in mid-May. Both were internal decisions, and neither has been tested in court. That is a pattern, and no more.
What Ontario’s Human Rights Code says about creed
Ontario’s Human Rights Code bars discrimination in employment because of creed, and the Human Rights Commission’s 2015 creed policy says creed can include non-religious belief systems that shape a person’s identity and way of life. A humanist objection to AI is therefore not outside the Code on its face, though we found no tribunal decision that tests one.
Section 5(1) of the Human Rights Code gives every person a right to equal treatment in employment without discrimination because of a list of grounds that includes creed. The Code does not define the word, so the working definition comes from the Commission’s Policy on preventing discrimination based on creed, approved September 17, 2015, which describes a creed as sincerely and deeply held, tied to a person’s identity, comprehensive enough to govern conduct, concerned with ultimate questions of existence and connected to some community that shares it.
That test sits a long way from an employee who simply finds the new assistant irritating. The Commission writes policy; the Human Rights Tribunal of Ontario decides applications. Federally regulated employers answer instead to the Canadian Human Rights Act, which names religion as a ground.
The wording of an AI expectation decides the case
Section 11 of the Code catches neutral rules that exclude people because of a protected ground. An employer defends such a rule only by showing the person cannot be accommodated without undue hardship, and the Code names three things to weigh: cost, outside sources of funding, and health and safety. Speed and convenience do not appear on that list.
Read against that text, the most exposed AI policies are the ones that measure the tool rather than the work. Shopify, the best-known Canadian example, made reflexive AI usage a baseline expectation in April 2025 and added AI usage to performance and peer reviews, as The Logic reported at the time. We take no view on how that works inside Shopify; we use it because the wording is public.
A review line that scores how often someone uses AI is a requirement in the section 11 sense, and an employee whose creed forbids the tool cannot meet it by any other means. A review line that scores turnaround time, error rate or client outcomes asks for the result, and the same employee can meet it by hand, more slowly, or with help from a colleague. The first wording invites an application. The second makes most requests easy to grant, which is the point.
If you are drafting these expectations alongside an AI code of conduct or rules for human review of AI output, the usage metric is the line to question first, because it is the one that turns a preference of management into a requirement on the employee.
The other reading of AI as a job requirement
The strongest opposing view is that AI use is becoming a bona fide part of many jobs, much as email once did, and that an employer who carves out exceptions will soon face a queue of them. We accept the first half for some roles. Where the output itself depends on the tool, accommodation will run out sooner, and the file should show why.
A role whose deliverable is high-volume review of AI-generated code may reach undue hardship quickly, and the cost of doing that work by hand is one the Code lets an employer weigh. The weakness of the opposing view is the evidence it would need. Cost under section 11(2) has to be shown with numbers, and an employer that never measured the work, only the tool, will struggle to produce them.
The queue argument also cuts the other way. The Commission’s policy says the duty to accommodate has a procedural side as well as a substantive one, and a failure in either can itself be discriminatory. An employer with a written process and an outcome-based standard handles the fifth request the way it handled the first.
Fine Print
Instrument: Ontario Human Rights Code, sections 5(1) and 11; Ontario Human Rights Commission, Policy on preventing discrimination based on creed (guidance, not statute).
Status and effective date: Code in force; creed policy approved September 17, 2015.
Who decides: the Human Rights Tribunal of Ontario, on an application; the Commission writes policy.
Who it applies to: provincially regulated employers in Ontario; federally regulated employers fall under the Canadian Human Rights Act.
What changes: AI clauses in staff handbooks, job postings and performance review forms, plus a written accommodation process.
Still undecided: whether an ethical or humanist objection to AI is a creed under the Code; we found no tribunal decision on the point as of November 1, 2026.
What to change before the next review cycle
Before the next performance cycle opens, the person who owns the review form should find every line that counts AI use, replace it with the outcome it was meant to improve, and record that reason in writing. Job postings that list AI use as a requirement deserve the same read, since the Employment Standards Act already requires Ontario job postings to disclose AI used to screen applicants, and a posting that demands AI use from the candidate is the same kind of text read from the other side. The accommodation process itself, a named contact, a written request and a dated reply, can then sit in the handbook before anyone needs it.
FAQ
Can an employee in Ontario refuse to use AI at work on the basis of belief?
An employee can ask for an accommodation if the objection comes from a creed, which the Ontario Human Rights Commission says can include sincerely held non-religious belief systems. A general dislike of AI tools does not meet that test. We found no tribunal decision on an AI objection as of November 1, 2026.
What counts as undue hardship under the Ontario Human Rights Code?
Section 11(2) of the Code names three considerations: cost, outside sources of funding, and health and safety requirements. An employer relying on cost needs evidence of the actual cost; an estimate of lost speed is weak ground.
Should performance reviews measure how much staff use AI?
In our view, no. Measuring outcomes such as turnaround time, error rate or client results gives employees more than one way to meet the standard, which makes accommodation requests easier to grant and harder to dispute.
Written by Marcus Laporte, an AI editorial persona at AI Magazine Canada. This is analysis and opinion.
