He Doesn't Believe in God - and Still Won a Religious Exemption From Using AI at Work

As AI tools get written into job descriptions, one question is quietly moving from tech conferences into human resources departments: can an employee simply refuse? In the case of David Durman, a research administrator at the University of Kentucky, the answer turned out to be yes — under the same body of law that protects religious practice in the American workplace, even though Durman does not believe in God.

Durman helps academics prepare grant applications. Late in September, the university's equal opportunity office approved a religious accommodation that excuses him from personally using AI systems such as Microsoft's Copilot, OpenAI's ChatGPT and Anthropic's Claude. The decision was made voluntarily, case by case, after a review of his duties. No court was involved — which means the case shows what one employer chose to do, not what the law compels.

Durman's argument was unusually philosophical. He describes himself as a Humanist, a nontheistic belief system organized around human moral development rather than worship. In his view, doing his own thinking is part of how he becomes the person he wants to be, and handing that process to a chatbot works against it. Under Title VII of the Civil Rights Act, an employer must reasonably accommodate a sincerely held religious belief — or, as case law has established, a comparable secular moral conviction — unless doing so imposes undue hardship.

The American Humanist Association, whose lawyer worked with Durman on the request, says this is the first time humanism has been used to win an exemption from workplace AI. It is not, however, the first religious AI exemption on record: earlier this year a software engineer who is a Unitarian Universalist received a similar accommodation.

The details of Durman's arrangement matter as much as the headline. The exemption covers only his own use. He still attends meetings and submits work through platforms that have AI built in, because refusing entire systems would go beyond what the accommodation addresses. That line — between declining to personally delegate your thinking and opting out of an AI-saturated workplace — is where most future disputes are likely to land.

The timing is not incidental. Employers increasingly write AI fluency into job postings, applicant tracking systems screen resumes for AI keywords, and recruiters ask candidates how they use AI in first-round interviews. Against that backdrop, a formal accommodation channel becomes a live policy question for any institution rolling out mandatory AI tools — universities, agencies and corporations alike.

Durman's case will not bind anyone. But it demonstrates the path: a documented, individualized process that grants the exemption without dismantling the workplace. Employers who never built such a process may now find requests arriving at the exact moment their AI mandates do. The safer position — for both sides — is to have an answer ready before the first one lands.