A Humanist Won a Religious Exemption From Using AI at Work: What It Means for Employers and Staff

A University of Kentucky administrator has won a religious accommodation that lets him avoid using AI at work, and his objection is not based on belief in a god. David Durman, a nontheistic Humanist, secured the exemption with help from the American Humanist Association (AHA), according to TechCrunch, and he is described as the first Humanist to do so.

It is a small case with a broad question behind it: if a company or institution tells staff to use AI, what happens to the people who object on conscience grounds? This post covers what was granted, the legal framework as reported, the earlier case that inspired it, and why neither amounts to a precedent. It sits alongside our coverage of the AI verification tax and other tensions in how AI is being rolled out at work.

What was granted (TL;DR)

What did Durman say and do?

Durman supports researchers applying for grants and handles compliance work. He told TechCrunch that humanism is "a collection of beliefs about the human experience and how humans interact with the world," centered on well-being and flourishing. He argues that regular AI use erodes the active mental effort through which people become moral actors: "The way we become moral actors for ourselves and for others is through the active application of our minds," he said, and diminishing that through AI use would hurt that ability.

In a statement published by the AHA, he also cited effects on livelihoods, environmental costs and the use of human creative work to build models. Those are the same objections many secular critics raise, now framed as a conscience claim.

He did not wait for a mandate. TechCrunch reports that colleagues and leadership recommended trying AI tools and that he acted early, in case the suggestion hardens into a requirement. Newsweek, citing the AHA and an interview Durman gave The Chronicle of Higher Education, says the university began encouraging employees to fold AI into their workflows and that he sought the accommodation preemptively.

What the university said

The University of Kentucky's spokesperson, Jay Blanton, told TechCrunch that the institution reviews religious accommodation requests individually, looking at the employee's job responsibilities and circumstances. "In this case, the university determined that an accommodation could be provided while allowing the employee to continue performing the essential functions of their position," he said. He added that as technology and job duties evolve, accommodations may be reviewed to make sure they stay reasonable and workable.

That last sentence matters. It signals the accommodation is conditional, not permanent. If AI becomes part of the essential functions of a grant-administration role, the same analysis could come out differently.

The earlier case: Erin Maus

Durman said he was inspired by Erin Maus, a Unitarian Universalist software engineer. According to press coverage, including Business Insider's reporting as summarized by AI Weekly, Maus requested an accommodation in the spring, citing ethical objections including environmental impact and concern about job losses, and her employer approved it. She reportedly writes and reviews code by hand.

Her case is the more surprising one because, as TechCrunch notes, it has become rare for software engineers to write code with no AI assistance. For a developer, a no-AI accommodation touches the core of the job. For an administrator whose AI use is optional, as Durman's was, the accommodation costs the employer far less. That difference is the key to understanding how far these cases can generalize.

What does Title VII actually require?

As described in the coverage, the request rests on Title VII of the Civil Rights Act of 1964, which bars employment discrimination based on religion. The standard employers apply is whether the belief is sincerely held and whether accommodating it would impose undue hardship on the business.

Three points follow from that framework, and they are general explanations rather than legal advice:

1. Religion is defined broadly. Protected belief systems are not limited to theistic faiths. The AHA's staff attorney, Alexandria Huiskens, told TechCrunch the group wants Humanists and other nontheists to know "they are also protected by the law at work."
2. Sincerity is the test, not popularity. An employer can ask whether the belief is genuinely held. A preference or a general dislike of a tool is different from a conscience claim.
3. Hardship depends on the job. An exemption that is easy to give for a role where AI is a convenience may be hard to give where AI is the core workflow.

There is no court ruling here. Both outcomes are voluntary employer decisions after a request, so a different employer could refuse and a court could rule either way.

How far does the exemption go?

Coverage of the AHA statement says Durman is excused from personal AI use but must still attend meetings where AI is used, work with colleagues who use it, and submit materials to AI tools when needed. That is an important limit. He is not opting the workplace out of AI. He is opting himself out of being the person who operates it.

This distinction will matter in practice. A worker who produces a document that colleagues then run through an AI summarizer is still part of an AI-touched process. The accommodation covers the first-person act of using a tool, not every interaction with AI-generated output.

Why this matters now

Several trends make these requests more likely.

• Mandates are spreading. Companies increasingly tie AI adoption to performance goals. Our coverage of Microsoft's buyout offer to about 60,000 employees shows how tightly AI strategy and headcount decisions are now linked.
• Worker anxiety is organized. Unions and tech worker groups are building campaigns, as in the Irish tech union drive tied to AI layoffs and the dispute over Amazon's AI-driven risk scores for warehouse workers.
• Institutions are drawing their own lines. Open-source projects now set explicit rules for AI contributions, as in the Haskell compiler's human-authorship policy. A conscience-based opt-out is the individual version of the same boundary.
• Religious voices are weighing in. TechCrunch notes that Pope Leo XIV has described AI as a threat to human dignity. Statements from religious leaders give workers vocabulary and a basis for sincerity claims.

Productivity data cuts the other way. Surveys such as Gallup's teacher study show many workers do adopt AI, and employers will argue that essential functions are changing. The tension between those two forces is exactly where accommodation cases will be decided.

What should employers do?

None of this is legal advice, but the practical shape is visible.

1. Write down essential functions. If AI use is truly essential to a role, say so, with specifics. If it is optional, say that too. Vague "everyone must use AI" language makes accommodation analysis harder.
2. Route objections to HR, not performance management. A religious or conscience objection is a request to be assessed, not an insubordination issue to discipline.
3. Assess hardship on the facts. Ask what it would cost to let this person do the job without AI: time, quality, risk. Document the answer.
4. Review periodically. Kentucky's own statement says accommodations may be revisited as roles change. That is a reasonable approach if it is applied fairly.
5. Do not assume every objection is religious. Many workers have non-religious concerns about AI. Those belong in separate channels such as training, consultation and policy design.

What should workers take from it?

The AHA's message is that nonreligious people should know the law may cover them too, and Durman said he hoped others with similar concerns would "pursue their concerns to the best of their ability where they feel is the most appropriate", not necessarily through a religious exemption. Practical points, again not legal advice:

• Check your employer's actual policy. Many AI "mandates" are encouragement, not requirement, as in Durman's case.
• A sincerely held belief is the foundation. A request made only to avoid a tool you dislike is unlikely to qualify.
• Put the request in writing and describe the specific tasks you object to, not AI in general.
• Consider where your job sits on the AI dependency spectrum. A role built around AI tools will be harder to accommodate than one where it is a convenience.

What we do not know

• How common these requests are. Durman is described as one of the first to discuss his publicly. Employers do not report accommodation counts.
• How other employers would rule. Two approvals do not tell us about refusals, which are less likely to be publicized.
• Whether a court would agree. No litigation has been reported in the coverage we reviewed.
• The exact documents. The AHA's original statement is the primary source for the accommodation terms. The details above come from press coverage of it.

Bottom line

Two workers have now persuaded employers that opting out of AI can be a matter of conscience, one in software and one in university administration. The cases are real, but they are individual and conditional, and they sit in a legal area without precedent for AI. For employers, the lesson is to define what "using AI" means in each role before mandating it. For workers, the lesson is that the first question is not whether you can opt out, but what your job actually requires.

Related reading

• The AI verification tax: rework surveys and how to cut it
• Microsoft buyouts for about 60,000 employees
• Irish tech union and AI layoffs
• Amazon Atlas AI risk scores for warehouse workers
• Haskell compiler AI contribution policy
• Gallup: teachers and AI
• Goldman Sachs on AI and worker displacement

Details are accurate as of October 30, 2026 and based on TechCrunch, Newsweek and other press coverage of the AHA statement. This post is general information, not legal advice.