Requests to opt out of workplace AI tools on religious grounds are beginning to reach HR departments, usually in unremarkable form: a short email explaining that an employee cannot use the company’s AI drafting tool because of their faith, in a role built around that tool. Pope Leo XIV’s May 2026 statement on AI has drawn attention to the subject, and Catholic, Latter-Day Saints, evangelical, Jewish, Islamic, and Seventh-Day Adventist bodies have each addressed the technology in some form. For employers, the accommodation decision is often the more straightforward part. The questions that follow tend to be less settled: how sincerity is assessed, how performance is measured once a tool is set aside, and how the remaining work is redistributed.
The framework, briefly
Title VII requires accommodation of a sincerely held religious belief absent undue hardship. After Groff v. DeJoy, 600 U.S. 447 (2023), an employer must show that granting the accommodation would result in substantial increased costs in relation to the conduct of its particular business, and the EEOC has made religious accommodation an express enforcement priority. Colleagues at FordHarrison and Ogletree Deakins recently covered this topic, including the mechanics of the interactive process and what does and does not qualify as hardship. Less attention has been paid to what comes afterward, once the legal standard has been applied and the remaining questions are factual ones.
Handling questions about sincerity
Employers rarely challenge sincerity, and caution is warranted when they do. Objections to AI may tempt them, because these requests often arrive dressed in secular clothing: environmental impact, job displacement, surveillance, or plain distrust of the tool. The gray space is not whether the question can be asked, but how far they can go.
When an employer decides to examine sincerity, the boundaries of that inquiry matter more than the conclusion. Internal or external, the investigative inquiry can fairly consider whether the stated belief is consistent with the employee’s observable conduct, when the request was made relative to any performance action, and whether what the employee is asking for follows from the belief itself. What should fall squarely outside of scope is the accuracy of the theology, whether others in the same faith agree, whether a member of the clergy would endorse the position, or how recently the employee arrived at the belief. Those boundaries are easier to hold when they are set out in writing before the inquiry begins, rather than worked out by whoever first received the request.
When the employee’s stated belief does not match official doctrine
Pope Leo counseled caution about AI. He did not tell Catholics to stop using it. Some employees are nonetheless reading a prohibition into what he wrote, and some Seventh-Day Adventist employees are raising a separate concern that AI takes on a role reserved to God.
It may be tempting to point out that the employee’s own church does not require what the employee says it requires. But well-versed HR and employment advisors recognize that this is exactly the kind of second-guessing Title VII does not permit. An unusual reading of religious authority, even a mistaken one, is still protected. This is worth keeping in mind when the scope of an inquiry is being drafted, since a question framed as whether the employee’s belief matches official doctrine is one that an investigator is not in a position to answer.
Sorting a religious objection from a personal one
Separating a religious objection from a personal preference is well-established ground for investigators, and AI objections are no exception. The origin and duration of the belief can be traced independent of the AI question. The employee can be asked to articulate the underlying religious principle without referring to AI at all, and the request can then be tested against that principle. An employee who declines AI for drafting but accepts it for calendar scheduling may be drawing a principled line or a self-serving one, and the answer lies in the reasoning rather than the pattern. It is also worth remembering that beliefs can read secular on the surface and religious underneath. An objection grounded in environmental stewardship as a tenet of faith is a clear example.
Check the exceptions already in place
The roughly $600,000 verdict in the biometric hand scanner case turned on an alternative the employer had already extended to employees with hand injuries. The same reasoning applies here. Most employers already maintain non-AI pathways on some level in the organization: privileged communications, client contracts restricting tool use, security-sensitive work, PHI handling, union carve-outs. Where one exception already exists, the argument that manual performance is not workable becomes harder to sustain. Taking stock of those exceptions early gives a clearer picture before undue hardship is raised.
When performance metrics assume AI use
AI is increasingly embedded in productivity measurement, not just workflow, and that deserves attention at the time an accommodation is granted. One sequence worth anticipating: the accommodation is approved, output targets are left untouched, the employee falls short of AI-benchmarked metrics, a performance improvement plan follows, and the employment relationship ends. What reaches HR may then be a retaliation or disparate treatment complaint rather than a failure to accommodate claim. The questions that tend to matter most are whether anyone revisited the metric when the accommodation was granted, who made that decision, and whether it was documented. That step may not appear on accommodation checklists, and it is worth adding.
What follows for the rest of the team
Coworker resentment does not defeat an accommodation after Groff, but it does not disappear either, and one employee stepping back from a tool may shift work onto colleagues. Two kinds of complaints can follow: religious harassment reported by the accommodated employee, who may be labeled a Luddite or a zealot, and preferential treatment concerns raised by teammates absorbing the redistributed work. Planning the team conversation at the time of the accommodation helps, as does explaining a workload shift without disclosing the religious basis for it.
Investigator’s own tools
A related question deserves attention within our own profession. If an investigator uses AI transcription or other AI tools, can an employee whose objection is to AI decline that process? Does an external investigator owe a witness a manual alternative? What about AI-assisted hotlines and ticketing systems where a complainant never reaches a human being? Our view is that investigators should disclose their AI use at intake and be prepared to offer an alternative, for the same reasons we advise employers to.
To date, appellate courts have yet to take up a religious objection to workplace AI. When these matters do surface, they may not be framed that way. They may arrive as performance disputes, harassment complaints, or sincerity questions that reach an investigator without a clear scope. Employers who settle questions of scope, performance expectations, and workload at the time of the accommodation are better positioned if a complaint follows.
If your organization is weighing a religious accommodation request involving AI or working through a complaint that has grown out of one, Aecus Law is available to help.