Workplace AI and the Right to Refuse: Progress Needs Worker Choice

Techstrong reported on October 29 that United Tech and Allied Workers is calling for legal protection for employees who refuse workplace AI. [1] UTAW published its demands on October 2, including negotiation over deployment and protection against penalties for refusal. These are the union's proposed protections. [2]

I support the ambition to expand human ability through technology. But I think workplace AI deserves to be called progress only when employees gain meaningful influence over how their new capabilities are used. Making someone faster while giving them less say over their work would be a rotten bargain.

The case for enhancement includes choice

Humanity+'s Transhumanist Declaration, adopted in 2009, supports extending human potential while emphasizing autonomy, individual rights and personal choice over enhancement. It also recognizes that technological change can cause serious harm. [3]

I think those commitments belong together. AI that helps someone understand unfamiliar material or express an idea could expand their practical freedom. I see a defensible case for that kind of enhancement, including at work.

That freedom becomes questionable if every new capability brings an obligation the employee cannot challenge. Applying the declaration to employment is my interpretation; the document does not itself prescribe workplace AI refusal rights. [3]

Faster work leaves the bargain unsettled

Gallup's October 6 report found that 63% of employees who had used AI at least a few times in their job during the previous year said it helped them work faster. Thirty-one percent said their AI use had prompted employers to request additional responsibilities. The surveys ran January 26–March 24, 2026. [4]

Separately, 52% of all surveyed employees wanted greater influence over new technology adoption. That question covered technology generally; it did not ask whether respondents supported an AI refusal right. These are different populations and self-reported assessments, and the findings do not establish causation. [4]

More responsibility might mean worthwhile development. It might also mean a bigger workload without a say in priorities. The findings leave room for both possibilities. My judgment is that a time saving should open a discussion about what happens next.

An objection needs a route to changing the decision

UTAW calls for consultation, negotiation and collective bargaining alongside protection for refusing AI. [2] I favor a process that lets employees challenge a tool without the objection itself becoming grounds for punishment.

I would want that process to begin before software is purchased or performance targets change. Someone must have authority to pause or revise deployment. Collecting feedback after every consequential decision has been made offers little influence.

Consider a hypothetical employer that expands AI from summarizing meetings into assessing individual performance. Agreement to the first purpose would tell us little about acceptance of the second. I would expect fresh consultation about the changed purpose and a way to dispute assessments before they affect career prospects.

The appeal of centralized decisions is real: someone owns the choice, sets a deadline and coordinates the team. I take that seriously. Yet a fast decision can preserve a bad assumption when the people affected cannot challenge it.

Refusal needs practical boundaries

I distinguish declining personal use from blocking an entire deployment. UTAW's short announcement does not resolve how conflicts between refusal and shared workflows would be handled. [2] That practical question deserves attention.

If colleagues must maintain incompatible processes, extra work could fall on people who welcomed the technology. A permanent exemption from every change would also make organizing a team difficult. Those are foreseeable problems, not measured outcomes here.

My preference is for specified grounds for objections, prompt review and decisions with reasons that workers can challenge. A quality concern might justify an alternative method or a pause to test the tool. An employer should have to explain why the disputed use is necessary.

I would still allow employers to require common tools when evidence supports their use and objections have received meaningful review. Protection should make criticism consequential without promising that every employee will always get their preferred decision.

Leaving cannot be the whole answer

For someone with no alternative job offer, resignation can be a costly way to object. I think treating departure as the sole answer to disagreement offers a thin version of freedom. The person still needs a way to influence the job they have.

For career planning, I would treat an employer's arrangements for employee input as part of job quality. Access to capable software matters. So does knowing how disputed targets or evaluations get reconsidered. I would give concrete examples of employees changing a rollout more weight than a promise of empowerment.

This is also a question of professional development. If workers are expected to exercise judgment, I think they need room to disagree about the tools shaping that judgment. A workplace that rewards unquestioning compliance could make someone more productive while giving them less practice making consequential decisions.

My expectation is that AI will produce different working lives under different management arrangements. It could expand discretion and remove tedious work. It could also compress more obligations into the day while narrowing room to object. These sources cannot establish which pattern will dominate.

The transhumanism I can defend gives people greater capability and greater control over its use. If employees gain the first while losing the second, I would call that a failure of the employment bargain.