For two years, AI in the workplace was an IT story and an efficiency story. As of mid-2026, it has become a labor organizing story — and the workers winning the clearest protections aren't the ones waiting for legislation, they're the ones at the bargaining table writing enforceable contract language while lawmakers debate and corporate executives experiment. The International Longshoremen's Association negotiated a complete prohibition on fully automated port technology. The Las Vegas Culinary Workers Union contractually obligated covered employers to bargain before implementing any AI decision. The Writers Guild of America secured a ban on AI-generated scripts that applies at the largest studios. And the Communications Workers of America has published an AI bargaining toolkit so that what wins at one company can be replicated across an entire industry.
This is the fifth piece in our real-life AI series second volume. This one focuses on the workplace power dynamic: how AI is changing the leverage that employers and workers each bring to the negotiating table, what unions have actually won in contracts, and what workers without union representation can realistically do in a workplace where AI is increasingly shaping how their job is measured, managed, and eventually made redundant.
What Workers and Employers Are Each Fighting For at the Table
| Issue | Employer Position | Worker/Union Position | Where Contracts Have Landed |
|---|---|---|---|
| Job displacement from AI automation | AI replaces tasks, not jobs; new roles will emerge; efficiency gains benefit the business | Displacement is real and concentrated in specific roles; workers need advance notice, retraining, and layoff protections | ILA: full ban on automated technology in ports. WGA: ban on AI writing scripts. APWU (postal): "no layoff" clause protecting workers from automation-driven cuts. |
| AI performance monitoring and surveillance | AI enables objective performance measurement; reduces supervisor bias; improves productivity tracking | Algorithmic management removes human judgment; creates inhumane performance pressure; data used against workers in discipline | UPS Teamsters 2023: management can't use sensor data as sole basis for discipline. Las Vegas Culinary: must bargain before implementing AI surveillance tools. |
| AI use in hiring and promotion | AI reduces hiring bias; enables consistent evaluation; speeds up large candidate pools | AI hiring tools have documented bias; workers have no recourse against algorithmic decisions; lack transparency about how decisions are made | NLRB has found that unilateral implementation of AI hiring tools without bargaining may be an unfair labor practice in unionized settings. |
| Training data and creative work | AI trained on existing company/creative work is a business asset; accelerates content production | Workers' existing creative output used as unpaid AI training data without consent or compensation | WGA: studios prohibited from using writers' material as AI training data. ZeniMax (Microsoft): consent and compensation required before using workers' work for AI training. |
| Wage sharing from AI productivity gains | Productivity gains justify maintaining or reducing headcount rather than raising wages | If AI makes workers more productive, workers should share in the value created, not just absorb the efficiency as profit | Limited wins so far; Partnership on AI case studies show this remains the hardest issue to move |
The Landmark Contracts That Set the Templates in 2025–2026
The WGA settlement is where this history effectively begins for the current era. The 2023 Writers Guild of America strike — one of the longest labor disputes in the history of US entertainment — explicitly demanded limits on AI for entertainment content production. The eventual agreement contractually prevents the use of LLMs to write scripts or to use AI output as source material, and prohibits the use of writers' material as training data. That contract established the template: specific prohibitions on specific AI uses, written into an enforceable collective agreement rather than left to policy statements or goodwill.
The ILA ports agreement went further. The International Longshoremen's Association negotiated a new provision that prohibits all fully automated technology at the ports it covers — a complete ban rather than a regulation. This is the most maximalist position any major union has achieved, and it came from a union that had watched automation displace dock workers at non-union ports and decided that managed introduction was insufficient protection.
The Las Vegas Culinary Workers Union took a different approach that's been described as a procedural model for other industries. Rather than banning specific AI tools, the agreement contractually obligates covered employers to bargain over any decision to implement AI in the workplace. This doesn't prevent AI adoption — it gives the union a seat at the table before it happens. Employers have to negotiate the terms before deployment rather than presenting AI as a fait accompli and managing the consequences afterward.
Partnership on AI's April 2026 case study of three agreements that secured AI protections for 30,000 union workers identified a key strategic pattern: unions should use each other's agreements as models. What ZeniMax workers won from Microsoft on AI consent and compensation became a template CWA used in negotiations at other tech companies. What the Culinary Workers won on mandatory pre-implementation bargaining became a reference point for service industry unions across different sectors.
The Employer Side of the Ledger: What Companies Are Actually Doing
It's worth being honest about both sides of this dynamic. Employers are not simply trying to harm workers — many of the AI tools being deployed do genuinely reduce tedious workloads, improve safety, and create real productivity gains that benefit the business. The dispute is almost never about whether AI should exist in the workplace. It's about who controls the terms of implementation, who bears the costs of disruption, and who captures the financial value of the productivity gains.
On the surveillance and performance management question specifically, employer arguments aren't without merit in some contexts. AI-powered performance monitoring can reduce the subjective bias of individual supervisors, provide more consistent standards, and catch patterns of unfair treatment that were previously invisible. The problem documented in labor research isn't that AI measurement exists — it's that algorithmic management creates performance pressure that humans have found genuinely difficult to sustain, that the measurement systems are often opaque (workers can't see how they're being scored or why), and that AI-generated performance data is used in discipline and termination decisions without human review of the underlying context.
The legal landscape is also genuinely uncertain for employers. Whether and how AI may trigger mandatory bargaining is a significant legal question for unionized employers. The National Labor Relations Board has found that unilateral implementation of AI tools that affect terms and conditions of employment without bargaining may constitute an unfair labor practice. That creates real legal risk for companies that implement AI workplace tools without consulting unions in unionized settings, and it's why many employers prefer negotiated agreements that provide legal clarity over unilateral implementation that creates exposure.
For Workers Without Union Representation: What Actually Exists
The majority of U.S. private-sector workers are not in unions — union membership sits around 6% in the private sector — which means the contract wins described above don't directly apply to most workers. What protections do non-union workers actually have?
The honest answer is: fewer, and less enforceable. Federal law (the National Labor Relations Act) protects the right to organize and engage in concerted action even without a formal union, which means non-union workers who collectively raise concerns about AI implementation to management have some legal protection against retaliation. But without a contract, there's no enforceable agreement about how AI will be used, how it will be introduced, or what happens if it affects their job.
State-level legislation is beginning to fill some gaps. Several states have introduced bills requiring employers to disclose when AI is used in hiring, performance evaluation, or termination decisions. Illinois enacted legislation specifically requiring disclosure of AI use in hiring. New York City passed a law requiring bias audits of AI hiring tools with results made public. As of mid-2026, federal AI workplace legislation remains stalled, leaving the same patchwork-of-states situation that characterizes data privacy law.
For individual non-union workers, the most actionable near-term steps are: asking HR directly what AI systems are used in your performance evaluation and what data they collect (you have a right to ask; the answer may be revealing), raising concerns collectively with colleagues rather than individually (concerted action has legal protection), connecting with worker advocacy organizations that track AI in workplaces, and following legislative developments in your state specifically around AI workplace disclosure requirements.
What This Means for the Near Future
The trajectory here is clearer than in most areas of AI labor policy. For a couple of years, AI in the workplace was an IT story and an efficiency story. Now it is a union organizing story. That shift has momentum — workers at previously non-union workplaces are organizing specifically because of AI concerns, existing unions are making AI protections a priority in bargaining, and the contract language being negotiated in 2025–2026 is becoming the template for the next generation of agreements.
The key variable is union density. The protections workers are winning — mandatory pre-implementation bargaining, prohibitions on specific uses, consent requirements for training data, no-layoff clauses — are all contract-based protections that require collective bargaining power to obtain. In the absence of legislation that extends equivalent protections to all workers regardless of union status, the gap between what union members can negotiate and what individual workers can protect themselves with is likely to widen as AI deployment accelerates.
Frequently Asked Questions
Do employers have to consult unions before implementing AI in the workplace?
In unionized settings, if the AI system affects terms and conditions of employment — wages, hours, working conditions, job security — the National Labor Relations Board has indicated that unilateral implementation without bargaining may be an unfair labor practice. The Las Vegas Culinary Workers Union secured a contract that explicitly requires pre-implementation bargaining for any AI decision, which goes further than general NLRB guidance.
Can AI monitoring data be used to discipline or fire workers?
In non-union workplaces, generally yes — employers have wide discretion to use AI performance data in employment decisions, subject to anti-discrimination law and any applicable state disclosure requirements. In unionized settings, specific contract language governs this — the UPS Teamsters contract, for example, prohibits using sensor data as the sole basis for discipline, requiring human review of context before AI-generated performance data is acted on.
What did the WGA strike actually win for writers on AI?
The WGA's agreement with the Alliance of Motion Picture and Television Producers prohibits studios from using AI to write scripts or to use AI-generated output as source material, and prohibits using writers' material as training data for AI without consent. This effectively prevents the studios from replacing human writers with AI or using writers' existing work to train a system that would eventually replace them.
What can non-union workers do about AI in their workplace?
Non-union workers have the right to raise AI concerns collectively with colleagues — concerted action is protected even without formal union status. They can ask HR directly what AI systems are used in their evaluation and what data is collected. They can connect with worker advocacy organizations tracking AI in workplaces. And they can follow state-level legislation on AI workplace disclosure, which is more active than federal law right now.
Are there examples of unions successfully blocking AI automation entirely?
Yes — the International Longshoremen's Association negotiated a complete prohibition on fully automated technology at the ports it covers, which is the most maximalist AI labor protection in any major U.S. collective bargaining agreement. The WGA also effectively banned AI from replacing human writers at covered studios. Both cases reflect industries where workers had sufficient bargaining power to achieve outright prohibitions rather than managed introduction.