Can You Refuse to Use AI at Work? Lessons for UK Employers from a US Religious Exemption Case

US programmer wins religious exemption from using AI – what UK employers should learn

A Reddit post claims a 34-year-old programmer in North Carolina has been granted a religious exemption from being required to use AI at work. The story, originally reported by Business Insider and summarised by Futurism, has kicked off a lively debate about employee rights and AI mandates.

The headline question for UK readers: could this happen here? Short answer – yes, albeit under different legal tests. And employers should act now to avoid messy disputes later.

Read the Reddit thread and Futurism’s write-up for context. Details are limited, but the discussion is timely.

What the US case involved (based on what’s disclosed)

The post describes a developer, Erin Mouse, who objected to using AI tools at work on environmental grounds linked to her Unitarian Universalist beliefs. Her employer granted an exemption under US employment law (Title VII of the Civil Rights Act).

“Programmer Erin Mouse received an official religious exemption from the employer.”

Key points from the thread and linked article:

• Reason for objection: ecological concerns about energy and water consumption in data centres.
• Work practice: she codes “by hand”, claiming similar speed to AI-assisted work.
• Policy signal: commentators say US employers will need clear policies to handle religious accommodation requests.
• Not disclosed: the employer’s name, the specific AI tools in question, how performance is measured, and whether the exemption is time-limited.

On the environmental angle, data centres can use significant water for evaporative cooling and substantial electricity for compute. The true impact varies by design, climate, and operator. For a grounded look at water cycles and measurement, see my explainer on AI, waste water and data centre cooling.

How UK law differs – and where it leads to similar outcomes

US law (Title VII) requires “reasonable accommodation” for religion unless it causes undue hardship for the employer. UK law takes a different route but can land in a similar place.

Equality Act 2010: religion or belief is protected

• The Equality Act 2010 prohibits discrimination on the basis of religion or belief. This includes some “philosophical beliefs” that meet legal tests (serious, cogent, worthy of respect in a democratic society).
• UK tribunals have recognised strong, coherent environmental beliefs as protected in principle. Each case turns on the facts.
• A blanket rule that “everyone must use AI tools” is a workplace requirement (often called a provision, criterion or practice). If it disadvantages people with a protected religion or belief, it can amount to indirect discrimination unless the employer shows it’s a proportionate means of achieving a legitimate aim.

In practice: if an employee objects to using AI on religious or protected philosophical grounds, an employer should consider alternatives before insisting. There’s no automatic right to opt-out, but there’s a duty to justify any mandate. See Equality Act 2010 and ACAS guidance on religion or belief discrimination.

Why this matters now for UK organisations rolling out AI

Many teams are moving from AI experiments to policy. As they do, they risk tripping over employee rights, confidentiality, and data protection in one go. Three considerations stand out:

1) Mandates vs. choice

An “AI-by-default” rule can be hard to justify if you haven’t measured the actual productivity and quality impact in your context. If you can meet business needs to the same standard without forcing a particular tool, tribunals may expect you to be flexible.

2) Environmental objections are not fringe

Objections grounded in environmental belief could be protected. A thoughtful process – not a reflex “use the AI or else” – is essential.

3) Data protection and confidentiality

Some AI tools transmit code or customer data to third-party providers for processing or model improvement. That brings UK GDPR obligations into play, including lawful basis, processor agreements, and international transfers. If staff are uneasy about sending sensitive information to external models, their concerns are not just moral – they may be compliance risks. Consider completing a DPIA (Data Protection Impact Assessment) for material AI use.

Practical steps for UK employers adopting AI without falling foul of the law

Build a clear AI use policy

• Define approved tools, permitted use cases, and banned inputs (e.g. secrets, personal data) with examples.
• Explain how outputs are verified to manage hallucinations (made-up facts) and bias.
• State a process for requesting an exemption on religion or belief grounds and how alternatives will be considered.

Document your justification

• Run small A/B pilots to quantify productivity, defect rates, and review effort with and without AI. Evidence helps justify a requirement where needed.
• If you mandate AI for specific tasks, record why alternatives are not viable or proportionate.

Offer workable alternatives where reasonable

• Allow manual workflows with equivalent quality gates (e.g. code review, tests, SLAs) if they meet delivery timelines.
• Focus on outcomes (quality, speed, security), not the specific tool used, unless there’s a clear business or compliance reason.

Tackle the environmental question directly

• Be transparent about the tools you use and their data centre footprints where known.
• Prefer vendors with published sustainability reporting and options to limit data retention.
• Read a balanced view on AI and water use: AI, data centres and the water cycle.

Tighten your compliance posture

• Sign processor agreements with AI vendors; review model training and data retention settings.
• Avoid feeding models source code or personal data unless contractual and technical controls are in place.
• Train teams on prompt hygiene and output verification.

For developers and team leads: where AI really helps (and where it doesn’t)

• Good fits: unit test scaffolding, boilerplate, migration scripts, docstrings, sanity checks. These reduce grunt work.
• Riskier fits: complex logic, cryptography, licensing-sensitive code, or anything with safety implications. Expect hallucinations and subtle bugs.
• Measure honestly: time saved vs. review overhead, defect density, and rework. If “manual is as fast”, that should show up in metrics.

So, can you refuse to use AI at work in the UK?

Sometimes – if a requirement to use AI would indirectly discriminate against a protected religion or belief, and the employer can’t justify that requirement as proportionate. There’s no blanket right to opt out, but there is a right to be taken seriously and assessed fairly.

For employers, the answer isn’t to ban AI or force it on everyone. It’s to define when AI use is truly necessary, provide alternatives where reasonable, and back decisions with evidence and clear process.

Sources and further reading

• Reddit discussion: A US programmer just won a religious exemption from being forced to use AI at work
• Futurism summary: Religious exemption from workplace AI
• Equality Act 2010 (legislation): legislation.gov.uk
• ACAS guidance: Religion or belief discrimination
• Explainer: AI, waste water and data centre cooling

This article is general information, not legal advice. If you’re facing a live dispute, speak to an employment lawyer.