Can I Refuse AI at Work for Professional or Ethical Reasons? A Practical Guide

The honest answer: it depends on why you are refusing

There is no general legal right to refuse to use AI tools that your employer has directed you to use — in the same way that there is no general right to refuse to use any workplace technology your employer introduces. However, the reasons behind a refusal matter enormously, and several legitimate grounds exist that can make a refusal lawful, protected, or at minimum difficult for an employer to act upon without explanation.

The distinction that matters most is between refusing on grounds of personal preference (unlikely to be protected) and refusing on grounds of professional obligation, safety, privacy rights, or conscience (more likely to have legal support or at least legitimate weight in a workplace dispute). This article addresses each category and the jurisdiction-specific factors that determine the outcome.

Professional and ethical obligations — the strongest grounds

For professionals in regulated occupations — lawyers, accountants, healthcare workers, financial advisers, engineers, architects — the strongest grounds for refusing or limiting AI use are professional duty obligations. These are not just preferences; they are legally enforceable obligations that the professional personally carries and that an employer's instruction cannot override.

For example, a lawyer directed to use an AI drafting tool to prepare client documents may legitimately refuse or require human review if the tool produces outputs that, without professional review, could constitute providing incorrect legal advice. The obligation to exercise independent professional judgment is a personal duty, not delegable to an algorithm. Law societies in Australia (Australian Solicitors' Conduct Rules 9.1), the UK (SRA Code of Conduct), and the US (Model Rules of Professional Conduct) all confirm that the responsible use of technology is a professional duty, and that errors produced by AI that affect clients are the lawyer's responsibility, not the tool's.

Similarly, healthcare workers directed to use AI clinical decision support tools are not obliged to implement AI recommendations without applying their own professional judgment. The duty of care to patients is personal and cannot be discharged by following an algorithm. The same applies to financial advisers whose duty to clients requires personalised advice based on individual circumstances — an AI-generated recommendation does not satisfy this duty without human review and professional sign-off.

If your employer is directing you to use AI in a way that would require you to provide output to clients or make decisions that you cannot professionally stand behind without adequate review, you have a legitimate professional basis to require that review step to remain part of your process. Document your concern in writing and raise it through your professional body's guidance or ethics hotline if needed.

Privacy and data concerns

If using an employer-mandated AI tool requires you to input personal information about clients, patients, or other individuals into a system that you believe does not comply with applicable privacy law, you have a legitimate concern that goes beyond personal preference. In Australia, the Privacy Act 1988 makes APP entities (and their employees acting on their behalf) responsible for the handling of personal information. If you are directed to use an AI tool that processes personal information in ways that breach the APPs — such as sending client data to an overseas vendor that cannot provide adequate protections — the fact that your employer directed you to do so does not eliminate the privacy risk or the organisation's liability.

Raising privacy concerns about AI tools with your employer's privacy officer or Data Protection Officer (in the EU/UK) is not refusing to work — it is raising a compliance concern that your employer is obliged to address. The Privacy Act in Australia and GDPR in the EU give individuals rights in relation to their own data, but also create obligations on organisations that employees can legitimately point to when their employer's AI practices appear non-compliant.

Workplace safety grounds

If an AI system creates a work health and safety risk — including psychosocial risks from AI-driven monitoring, unpredictable AI scheduling that creates excessive stress, or AI tools that produce errors with safety consequences in high-stakes environments — employees can raise these concerns through WHS channels. In Australia, the model Work Health and Safety Act gives workers the right to cease or refuse unsafe work without penalty if they have a reasonable concern about an immediate serious risk. While applying this to AI tools requires the risk to be direct and immediate, the more common pathway is raising a formal concern with a health and safety representative (HSR) or through the employer's WHS management system.

What happens if you refuse without adequate grounds

Refusing to use an AI tool purely because you dislike the technology, are uncomfortable with change, or prefer your existing workflow is unlikely to be protected by law and may constitute a failure to follow a lawful and reasonable workplace direction. An employer who introduces AI tools to improve efficiency or quality, provides adequate training, and addresses legitimate concerns raised by employees is generally on solid legal ground in requiring employees to use those tools as a condition of employment.

The case law that will shape this area is still developing. A 2025 class action in the US (Mobley v. Workday) is testing whether employers are liable for discriminatory outcomes of AI hiring tools — but this addresses employer liability, not employee rights to refuse. The clearest employee-side protection currently comes from collective bargaining: in unionised workplaces, especially in Australia, the US, and the EU, enterprise agreements and collective agreements increasingly include provisions requiring consultation before AI is introduced to replace or significantly change work processes. If you are in a unionised workplace, your union delegate is the first call.

The approach most likely to work

Rather than framing the situation as refusal, the approach most likely to produce a workable outcome is: raise the specific concern in writing, reference the specific professional, safety, or privacy obligation that creates the concern, propose a modified approach (such as a required human review step), and document the exchange. This positions you as raising a legitimate compliance or professional concern rather than resisting change, which is a very different conversation for your employer to manage.