Religious Exemptions From AI at Work: The University of Kentucky Case
TL;DR
- University of Kentucky administrator David Durman reportedly secured a workplace exemption from using artificial intelligence while not identifying as a believer in a traditional god.
- The dispute tests how U.S. law defines religion, including whether sincerely held ethical or philosophical convictions can receive protection under workplace-discrimination rules.
- The case also raises practical questions about AI mandates, equal treatment, privacy, job performance and how employers should evaluate nontraditional religious objections.
The controversy at the University of Kentucky
A workplace dispute involving David Durman, an administrator at the University of Kentucky, has drawn attention to an unusual question: Can an employee object to using artificial intelligence on religious grounds without believing in a deity?
Reports about Durman’s situation describe him as having obtained an exemption from using AI in his work after presenting the objection as a religious belief. The reported arrangement has prompted debate because Durman apparently does not profess belief in a conventional god, challenging the assumption that a religious accommodation necessarily involves worship or theism.
The episode comes as universities, businesses and government agencies rapidly introduce generative AI tools for drafting documents, analyzing information, summarizing meetings and automating routine administrative work. Employers are increasingly moving from voluntary experimentation to formal expectations that workers use approved AI systems.
That shift creates a new category of workplace conflict. Employees may object to AI because of privacy concerns, fears about inaccurate or discriminatory outputs, professional ethics, environmental costs or opposition to replacing human judgment. Whether those objections qualify for legal protection depends on how they are characterized and whether they meet the definition of a sincerely held religious belief.
What is known about Durman’s objection
Public reporting has characterized Durman as a University of Kentucky administrator who sought permission not to use AI in the workplace. The precise terms of the accommodation, the technology covered and the administrative process used to approve it are important details, but not all have been publicly disclosed.
The available accounts do not establish that Durman received a universal right to refuse every form of workplace automation. An exemption may instead apply to particular tools, tasks or circumstances. Employers commonly tailor accommodations rather than issue blanket waivers, especially when an employee’s duties involve sensitive data, legal obligations or essential job functions.
The reports also do not necessarily mean that the university accepted a general proposition that opposition to AI is religious. An employer may approve an individual accommodation after assessing the employee’s sincerity and the effect on operations without declaring that every similar objection is protected.
That distinction matters. A single accommodation does not automatically create a policy that anyone can invoke to avoid AI-related assignments. It may reflect an individualized decision under the institution’s civil-rights procedures.
How U.S. law defines religion
Title VII of the Civil Rights Act of 1964 requires covered employers to accommodate an employee’s religious observance or practice unless doing so would impose an undue hardship on the employer’s business.
The law’s definition of religion is broader than membership in a church or belief in a deity. Courts and federal guidance have recognized that religious beliefs can be unconventional, nontraditional or shared by relatively few people. At the same time, not every personal preference, political position or moral concern qualifies as religion.
The central questions are typically whether the belief is religious in nature, whether the employee sincerely holds it and whether the requested accommodation is reasonable. Employers generally should not decide whether a belief is theologically correct. They may, however, examine whether a claimed belief is being asserted for an improper purpose or whether accommodating it would create operational problems.
This legal framework has made room for certain nontheistic belief systems. Humanism, ethical movements and other philosophies can sometimes be protected when they occupy a role in a person’s life comparable to that of traditional religion. But protection is fact-specific. A generalized objection to a new technology may look like a personal or professional opinion unless the employee can show that it arises from a broader system of religious belief.
The Supreme Court’s undue-hardship standard
The legal landscape changed significantly after the Supreme Court’s 2023 decision in Groff v. DeJoy. The Court rejected the previously common understanding that an employer could deny a religious accommodation whenever it created more than a minimal cost.
Under Groff, an employer must show that granting the accommodation would result in substantially increased costs in relation to the conduct of the particular business. The decision did not require employers to accept every requested accommodation, but it raised the level of justification needed to deny one.
For an AI-related request, an employer might argue that an exemption would undermine essential job duties, increase workloads for colleagues, create security risks, compromise service quality or prevent compliance with institutional rules. Whether those effects amount to undue hardship would depend on the facts.
A university could also consider alternatives. An employee might be allowed to complete a task manually, use a non-AI software feature, transfer a limited assignment to another worker or avoid uploading confidential information into an AI system. The least disruptive solution may be an accommodation that changes the method of performing work rather than eliminating the responsibility itself.
Why AI creates unusual accommodation problems
AI differs from many earlier workplace technologies because employees may object not only to the tool but also to the values and consequences they associate with it.
Some workers view generative AI as undermining human authorship or professional judgment. Others object to systems trained on copyrighted material, the collection of personal data, the environmental cost of large-scale computing or the possibility that automation will displace workers.
There are also concerns about reliability. AI systems can produce fabricated information, reproduce bias and conceal uncertainty behind fluent language. In fields such as education, health care, law and public administration, an employee may believe that delegating certain decisions to an algorithm violates professional duties.
Those concerns can be serious without being religious. That is one of the challenges raised by the Durman case: an employer must distinguish between a legally protected religious practice and a secular objection to technology while avoiding intrusive judgments about an employee’s worldview.
The difference between religion and conscience
The First Amendment protects religious exercise, but it does not give employees a general right to refuse any task that conflicts with their conscience. Workplace protections can extend beyond theistic faith, yet they do not turn every ethical conviction into a religious claim.
An employee who says, “I believe AI is harmful,” may be expressing a political, social or professional position. An employee who says, “My comprehensive ethical belief system forbids participation in practices that replace human moral agency,” may be presenting a claim that requires a more careful inquiry.
The wording alone is not decisive. Employers should look at the employee’s explanation, the consistency of the belief, how it functions in the person’s life and whether the request appears genuine. They should not demand proof of belief in God or require membership in an organized religion.
At the same time, employees cannot simply relabel a workplace preference as religion whenever an unpopular assignment arises. Sincerity does not guarantee an accommodation, and a religious label does not eliminate the requirement that the requested adjustment be workable.
What the university’s decision could mean
If the reports are accurate, the University of Kentucky’s decision illustrates one possible way institutions may handle AI-related objections: evaluate them individually rather than adopt a universal rule.
That approach has advantages. It allows administrators to consider the employee’s actual duties, the nature of the belief and the availability of alternatives. It also avoids declaring that all opposition to AI is either religious or secular.
But case-by-case decisions can produce uncertainty. Other employees may ask why one person received an exemption while another did not. Managers may struggle to apply standards consistently. Colleagues may be required to absorb tasks that an exempt employee declines, creating resentment or claims of unfair treatment.
Universities face an additional complication because they are both employers and educational institutions. They must balance workplace accommodation with academic freedom, research policies, student privacy, public-records requirements and institutional commitments to responsible technology use.
The limits of an AI exemption
An exemption from using AI would not necessarily excuse an employee from meeting performance standards. A worker may be permitted to complete an assignment without AI but still be required to meet deadlines, maintain accuracy and produce the expected work.
Nor would an exemption necessarily protect an employee who uses unauthorized tools. An institution can restrict access to confidential student, personnel or research information, regardless of whether the employee has a religious objection to AI. Data-security rules and legal obligations may apply to everyone.
The university could also require human review of AI-generated materials, prohibit automated decisions in sensitive contexts or limit AI use to approved platforms. In some situations, the accommodation may involve avoiding generative AI while allowing ordinary software automation that does not generate content or make judgments.
The broader policy question
The case arrives at a moment when employers are still writing basic AI policies. Many policies tell workers what tools they may use but say little about what happens when an employee declines to use them.
A durable policy should define which AI activities are mandatory, which are optional and which are prohibited. It should explain how workers can request an accommodation, identify who reviews requests and protect confidentiality during the process.
Employers should also avoid treating AI fluency as an end in itself. If the real objective is faster document preparation, better analysis or improved public service, managers should be able to consider non-AI methods that achieve the same result.
At the same time, organizations need clear limits. An exemption should not permit an employee to expose colleagues’ data, shift unreasonable work onto others or bypass a process that is legally required. Accommodations work best when they preserve the essential function of the job while changing how it is performed.
A test for future workplaces
Durman’s reported exemption may become an early example of a much larger category of disputes. As AI becomes embedded in scheduling, recruiting, customer service, software development and government administration, employees will increasingly challenge not only whether they must use it but also whether they must participate in systems they believe are morally wrong.
The legal answer will not turn solely on whether the employee believes in a god. Nor will it turn solely on whether the employer considers AI useful. The decisive issues are likely to be the sincerity and nature of the belief, the duties of the job, the availability of alternatives and the actual cost of accommodation.
For employers, the lesson is to prepare before disputes arise. AI policies should be coordinated with religious-accommodation procedures, disability accommodations, privacy rules and existing professional-ethics requirements.
For employees, the case suggests that a request should explain the underlying belief and propose a workable alternative rather than simply reject new technology. A credible accommodation is more likely to succeed when it protects the employee’s convictions without preventing the organization from carrying out its essential work.
The University of Kentucky episode therefore reaches beyond one administrator and one institution. It asks whether the modern workplace can require participation in an increasingly powerful technology—and how far the law should go when that participation conflicts with a worker’s deepest beliefs.
Get All The Latest Updates Delivered Straight To Your Inbox For Free!