Redistributed

When AI enters the workplace, who decides how it's used? For most workers, the answer is simple: management decides, workers adapt or leave. But for unionized workers, a different path exists. Collective bargaining — the process by which workers negotiate with employers as a group — is becoming a critical mechanism for determining how AI's gains are distributed.

This is redistribution through collective power. And it's happening now.

Why Unions Are Engaging with AI

For decades, unions focused on traditional concerns: wages, hours, safety, job security. Automation was a background threat, occasionally addressed in contract negotiations but rarely at the center.

That changed around 2023. Several factors converged:

AI became visible. Generative AI tools moved from research labs to workplaces almost overnight. Writers watched ChatGPT produce passable first drafts. Artists saw Midjourney generate images in seconds. Call center workers found AI handling entire conversations. The abstraction "automation might affect jobs someday" became "this system is doing parts of my job right now." The creative industries and software development were among the first to feel the impact directly.

Specific occupations faced immediate displacement. Unlike previous automation waves that affected manufacturing or routine clerical work, generative AI targets knowledge workers — writers, programmers, designers, customer service agents, translators. These workers are often organized and articulate. They could name what was happening to them.

High-profile strikes put AI on the public agenda. The 2023 Hollywood strikes — both the Writers Guild of America (WGA) and the Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) — centered AI provisions alongside traditional compensation issues. For months, entertainment industry labor disputes dominated American news. The AI clauses drew particular attention.

Legal frameworks lagged. Workers and unions realized that existing labor law offered little protection against AI-driven displacement. Unlike safety regulations or wage and hour laws, there was no requirement for employers to disclose automation plans, consult workers before implementation, or share productivity gains. If workers wanted protection, they would have to bargain for it.

What Unions Are Demanding

Across sectors and countries, a common set of demands is emerging. These represent a theory of how workers can retain power as AI reshapes work:

1. Disclosure and Transparency

Workers can't bargain over what they don't know about. The first demand is simple: tell us what you're planning.

This includes:

Advance notice of AI implementation decisions
Documentation of how AI systems work and what data they use
Impact assessments before deployment, covering job displacement, task changes, and working conditions
Ongoing reporting on AI's effects after implementation

The UK's Trades Union Congress (TUC) has been particularly focused on transparency, publishing research on "AI in the workplace" that emphasizes workers' right to understand systems that affect them (TUC, 2021). The TUC argues that without transparency, meaningful consultation is impossible — workers can't negotiate over black boxes.

Beyond disclosure, unions demand voice in decisions. Not just notification, but actual consultation before AI is deployed.

This takes different forms:

Consultation rights requiring management to discuss plans before implementation
Consent requirements giving workers or unions veto power over specific applications
Joint committees with ongoing oversight of AI systems
Pilot programs with evaluation before full rollout

European Works Councils, established under EU law, already have consultation rights on "substantial changes" to work organization. Unions are arguing that AI implementation constitutes such a change — and some are winning that argument (Countouris & De Stefano, 2024).

3. Retraining and Transition

If AI changes what workers do, employers should help workers adapt. Unions are bargaining for:

Retraining programs funded by employers, during work hours
Transition time before roles are eliminated or substantially changed
Internal mobility giving affected workers first claim on new positions
Severance packages when displacement can't be avoided

The logic is redistribution: if AI creates efficiency gains, some of those gains should fund worker transitions rather than flowing entirely to shareholders.

4. Job Guarantees and Employment Floors

The strongest protections limit displacement directly:

No-layoff clauses preventing job losses due to automation during the contract term
Minimum staffing levels that technology can't reduce
Human-in-the-loop requirements mandating human workers in specific functions
Sunset provisions requiring renegotiation if AI's impact exceeds projections

These provisions are harder to win than transparency or retraining, but they're being negotiated — particularly in sectors where unions have significant leverage.

5. AI-Specific Contract Language

Perhaps most importantly, unions are developing new contract language specifically for AI:

Definitions of what counts as AI or automation under the contract
Trigger clauses that activate protections when AI is introduced
Reopener provisions allowing mid-contract renegotiation if AI deployment accelerates
Ownership clauses addressing who controls AI-generated work product

The WGA and SAG-AFTRA strikes produced some of the most sophisticated AI contract language to date, creating templates other unions are adapting.

Case Study: The Hollywood Strikes

The 2023 Hollywood strikes represent the most visible — and arguably most successful — labor action over AI to date.

Writers Guild of America

The WGA strike began May 2, 2023, and lasted 148 days. AI was one of several issues, alongside streaming residuals and minimum staffing requirements. But AI provisions drew extraordinary public attention.

What writers feared:

Studios using AI to generate scripts or outlines, then hiring writers to "polish" them at lower rates. This wasn't hypothetical — some writers reported being asked to rewrite AI-generated material. The economics were clear: if AI produces first drafts, writers become editors, and their leverage collapses.

The WGA also worried about training data. AI systems learn from existing scripts. Studios own vast libraries of prior work. Without protections, writers' past labor would train systems that replace future writers.

What writers won:

The September 2023 agreement includes specific AI provisions:

AI cannot be credited as a writer. Only humans can receive "written by" or "screenplay by" credit.
AI-generated material is not "source material." If a studio provides AI-generated text to a writer, it doesn't reduce the writer's credit or compensation.
Writers cannot be required to use AI. Studios can't mandate that writers incorporate AI tools.
Studios must disclose AI use. If material given to a writer was AI-generated, studios must say so.
Training data protections. Studios cannot use covered work to train AI without consent.

The provisions aren't absolute — studios can still use AI in production, and writers can voluntarily use AI tools if permitted. But the agreement establishes that writers control how AI intersects with their work.

SAG-AFTRA

The actors' strike began July 14, 2023, and lasted 118 days. AI concerns were even more central than in the WGA dispute.

What actors feared:

"Digital replicas" — using an actor's likeness, voice, or performance without ongoing consent or compensation. Actors described being asked to submit to full-body scans, with minimal disclosure about how those scans would be used. The specter was clear: studios could capture a performer once, then generate unlimited performances without payment.

Background actors faced particular risk. Studios proposed capturing background performers' likenesses for a single day's pay, then using those digital extras indefinitely.

What actors won:

The November 2023 agreement addresses digital replicas extensively:

Consent required for digital replicas. Studios cannot create or use AI-generated replicas of performers without consent for each specific use.
Compensation required. Using a digital replica requires payment comparable to live performance.
Estate protections. Consent requirements extend to deceased performers, with estates retaining rights.
Background performer protections. Digital replicas of background performers cannot replace hiring actual background performers in scenes.
Synthetic performers. The contract establishes that fully AI-generated performers do not displace work that would go to human actors.

Like the WGA agreement, SAG-AFTRA's contract doesn't ban AI — but it ensures performers control their digital likenesses and receive compensation for their use.

Lessons from Hollywood

The strikes demonstrated several things:

Leverage matters. Hollywood unions have significant power — production schedules are tight, talent is recognizable, strikes create visible disruption. Not all workers have equivalent leverage. But the lesson is not that only Hollywood can win — it is that leverage must be built, and different sectors will build it differently.

Specificity matters. Vague concerns about "AI might take jobs" became concrete contract language. This required understanding both the technology and the industry's economics.

Solidarity matters. Both guilds struck simultaneously, despite different contract cycles. The combined pressure — no new scripts, no actors — was more effective than either alone.

Public sympathy can be won. The strikes received substantial public support, partly because writers and actors could articulate specific, relatable AI concerns. They weren't anti-technology — they were pro-worker-control.

Case Study: European Works Councils and Co-Determination

The Hollywood model centers on strike leverage and collective bargaining. Europe offers a different approach: legally mandated worker voice in corporate decision-making.

German Co-Determination

Germany's system of "Mitbestimmung" (co-determination) gives workers structural power over workplace decisions. Two mechanisms matter for AI:

Works councils (Betriebsräte): In companies with more than five employees, workers can elect councils with legal rights to information and consultation on workplace changes. Section 87 of the Works Constitution Act gives works councils co-determination rights over technical monitoring devices — language increasingly applied to AI surveillance and algorithmic management.

Board representation: In large companies (over 500 employees), workers elect representatives to supervisory boards. This gives labor voice in strategic decisions, including technology investments.

AI-specific applications:

German works councils have negotiated detailed AI agreements covering:

Prior notice and consultation before AI implementation
Limits on algorithmic surveillance
Requirements for human review of automated decisions
Data access and algorithmic transparency
Training and transition support

The IG Metall union has published model agreements specifically for AI in manufacturing, establishing templates other works councils can adapt (IG Metall, 2022).

European Works Councils

The EU's European Works Councils Directive requires multinational companies operating in Europe to establish councils representing workers across multiple countries. These councils have consultation rights on "transnational" matters — which increasingly includes AI deployment affecting multiple sites.

Advantages:

Cross-border coordination is essential for AI, which is often deployed as enterprise-wide systems. A decision made at headquarters affects workers in multiple countries. European Works Councils create forums for collective response.

Limitations:

Consultation is not consent. Employers must inform and discuss, but can proceed over worker objections. The process creates delay and forces explicit consideration of worker concerns, but doesn't provide veto power.

EU Platform Work Directive

The 2024 Platform Work Directive represents the EU's most direct intervention in AI workplace governance. Though focused on gig economy platforms, its principles extend further:

Presumption of employment for workers meeting specific criteria, addressing misclassification
Transparency requirements for algorithmic management, including information about how algorithms make decisions
Human review of significant automated decisions affecting workers
Consultation rights over algorithm design and implementation

The directive establishes that algorithmic management is a proper subject for labor law, not just a technical matter for employers to decide unilaterally.

Case Study: Warehouse and Logistics Organizing

Amazon warehouses represent algorithmic management at its most developed — and worker organizing in response.

The Amazon Labor Union

The Amazon Labor Union's 2022 victory at the JFK8 facility in Staten Island was the first successful union election at an Amazon warehouse in the United States. Algorithmic management was central to worker grievances.

What workers faced:

Workers described being managed by systems that tracked their every movement. "Rate" — items processed per hour — was calculated continuously. "Time off task" was monitored in minutes. Workers who fell below targets received automated warnings; repeated failures led to termination.

The system was opaque. Workers didn't know exactly how rate targets were set, why some tasks were harder than others, or how algorithms allocated work. Appeals went to managers who often deferred to the system's calculations.

What workers demanded:

The ALU campaign emphasized dignity and respect alongside traditional economic issues. AI-specific demands included:

Slower, sustainable pace allowing bathroom breaks
Transparency about how targets are calculated
Human review before termination for productivity reasons
Limits on surveillance

The challenge:

Amazon's intense opposition — including mandatory anti-union meetings, surveillance of organizers, and legal challenges — has slowed subsequent organizing. As of early 2026, JFK8 workers still lack a contract. But the campaign demonstrated that warehouse workers can organize despite algorithmic obstacles.

UNITE HERE and Hospitality

UNITE HERE, representing hospitality workers, has bargained over technology in hotels and food service. Their approach combines transparency demands with job security provisions.

In contracts with major hotel chains, UNITE HERE has won:

Advance notice before implementing guest-facing AI (chatbots, automated check-in)
Training programs when technology changes job requirements
Limits on surveillance of housekeeping workers
Consultation before adopting tracking or scheduling algorithms

The union's leverage comes partly from tight labor markets in hospitality — hotels need workers, and disruptive strikes during peak seasons create significant pressure.

SEIU and Public Sector

The Service Employees International Union (SEIU) represents many public sector workers who face AI in government systems — from benefit determination to fraud detection to service delivery.

SEIU locals have bargained over:

Algorithmic impact assessments for new government AI systems
Worker training before implementation
Guarantees that AI augments rather than replaces public sector jobs
Protections for workers who raise concerns about algorithmic errors

Public sector bargaining has a distinct advantage: government employers often have legal mandates for transparency that private employers lack.

Contract Provisions: What's Been Won So Far

Beyond the major cases, what AI-related contract language are unions winning?

Communication Workers of America (CWA)

The CWA represents call center workers, telecommunications technicians, and increasingly tech workers. Recent contracts include:

AI impact bargaining: Employers must negotiate over AI deployment and its effects
Disclosure requirements: 90-day notice before implementing AI systems
Training provisions: Employer-funded retraining for workers whose jobs are affected
Human contact guarantees: Customers must be able to reach human workers, not just AI

The CWA has also organized workers at Google, Microsoft, and other tech companies, creating the potential for organizing among AI developers themselves.

International Brotherhood of Electrical Workers (IBEW)

Representing utility workers, IBEW contracts often address automation in energy infrastructure:

Apprenticeship protections: AI cannot replace required human supervision of apprentices
Safety requirements: Human workers required for safety-critical functions
Scheduling transparency: Algorithmic scheduling subject to negotiation

American Federation of Teachers (AFT) and National Education Association (NEA)

Teacher unions are addressing AI in education:

Human grading requirements: AI can assist but not replace teacher evaluation
Surveillance limits: Restrictions on AI monitoring of teachers
Professional judgment: Teachers retain authority over pedagogical decisions, regardless of AI recommendations

Journalism Unions

NewsGuild, Journalists Union, and other media unions are negotiating:

Attribution requirements: AI-generated content must be disclosed
Job protections: AI cannot replace credited journalist positions
Training data rights: Journalists' work cannot train AI without consent

Limits of Bargaining

Collective bargaining has produced significant wins. But it has structural limitations.

Coverage Gaps

Only about 10% of U.S. private sector workers belong to unions. In the UK, union density is around 23%. Even in Germany, with its strong co-determination system, not all workers are covered.

The workers most exposed to AI — gig workers, contract workers, workers in non-union tech companies — often have no collective bargaining rights. Amazon warehouse workers might win contracts eventually, but DoorDash drivers, Uber drivers, and Instacart shoppers have no clear path to bargaining.

This creates a two-tier system. Unionized workers negotiate AI protections; non-union workers get whatever employers choose to provide. The freelance workforce is particularly exposed — bearing the sharpest edge of AI displacement with essentially no collective voice.

Scope Limitations

Even strong unions bargain at the firm or industry level. They can't address economy-wide AI deployment, tax policy favoring automation over labor, or research funding priorities.

Aaron Benanav's critique applies here: bargaining addresses how AI is deployed in specific workplaces, but not whether society should be organized around wage labor at all (Benanav, 2020). If AI truly transforms the nature of work, contracts that protect existing jobs may not be sufficient.

Employer Resistance

Companies invest heavily in defeating unionization. Amazon's anti-union campaign at JFK8 was legal — mandatory meetings, supervisor conversations, extensive communications — but intense. Subsequent elections at other Amazon facilities have mostly failed.

Tech companies have responded to organizing with layoffs that happen to affect union supporters, office closures at organizing sites, and legal challenges to labor board decisions. The resources are asymmetric.

Enforcement Challenges

Contract language is only as good as its enforcement. Workers report that employers sometimes violate AI provisions — implementing systems without required consultation, using data inappropriately, failing to provide promised training. Grievance procedures are slow. Arbitration is expensive. The labor board is backlogged.

International Arbitrage

AI systems are developed and deployed globally. A union that wins protections in one country may find the work shifted elsewhere. This is the same dynamic that affected manufacturing — but AI makes it easier, since work can move without physical relocation.

Policy Support: Beyond Bargaining

Because bargaining alone can't protect all workers, unions are advocating for policy changes that establish baseline protections.

Mandatory Consultation

Some jurisdictions are requiring employer consultation with workers before AI deployment:

EU AI Act: Requires consultation with worker representatives on workplace AI
EU Platform Work Directive: Mandates transparency and human review for algorithmic management
New York City Local Law 144: Requires bias audits for automated employment decision tools

These policies create floors that apply regardless of union status.

Algorithmic Transparency Laws

Workers can't bargain effectively without understanding what they're bargaining over. Transparency laws help:

California Consumer Privacy Act: Workers can request information about data collected about them
Illinois AI Video Interview Act: Requires disclosure and consent for AI in video interviews
EU GDPR Article 22: Rights related to automated decision-making

The TUC has proposed an "Accountability for Algorithms Act" that would require impact assessments, human review, and worker consultation for workplace AI (TUC, 2021).

Worker Classification

Many AI-affected workers are misclassified as contractors, denying them bargaining rights entirely. Policy to address misclassification — like California's AB5, the EU Platform Work Directive, or proposed federal standards — expands who can bargain.

Sectoral Bargaining

Some labor advocates propose sectoral bargaining — negotiating standards that apply across entire industries, not just individual employers. This would address coverage gaps, since even non-union workers would benefit from negotiated terms.

Countries like France and Germany have sectoral bargaining systems. Proposals to extend similar approaches to the U.S. (like the PRO Act's provisions for sectoral agreements) could transform AI labor negotiations.

Labor Board Enforcement

The National Labor Relations Board has signaled interest in algorithmic management as an unfair labor practice issue — particularly surveillance that chills organizing. Robust enforcement would make AI governance a labor rights matter, not just a bargaining subject.

The Redistribution Lens

Collective bargaining is a redistribution mechanism. It transfers power from employers to workers, giving workers voice in decisions that affect them. When that voice extends to AI deployment, it shapes how AI's productivity gains are distributed.

Without worker voice: Employers decide how to deploy AI, capturing efficiency gains while workers bear displacement costs. AI becomes a tool for extracting more output from fewer workers.

With worker voice: Workers participate in deciding how AI is used, ensuring that gains are shared through wages, benefits, job security, and working conditions. AI becomes a subject of negotiation, not an employer prerogative.

The mechanism matters. Bargained outcomes aren't just outcomes — they're expressions of worker power. The process of bargaining builds solidarity, develops expertise, and creates structures for ongoing engagement.

But worker voice in AI depends on worker power more broadly. Unions can only bargain what they have leverage to win. The Hollywood writers and actors had leverage; Amazon warehouse workers have less. The gains from AI bargaining will be unevenly distributed unless worker power expands.

This is why policy matters. Mandatory consultation, transparency requirements, sectoral bargaining — these policies don't replace collective action, but they change the terrain on which collective action happens. They establish that workers have legitimate interests in AI decisions, even when they lack the power to win protections through bargaining alone.

What Workers Can Do

Individual workers and workplace groups can take action even without formal union contracts:

Document AI in your workplace. What systems are used? How do they affect work? What information do workers have versus employers? Documentation is the foundation for any collective response.

Talk with coworkers. AI concerns are often shared. Informal conversations build the foundation for collective action — whether that means formal unionization, coordinated complaints, or simply shared understanding.

Use existing channels. Safety committees, employee resource groups, and even suggestion boxes can raise AI concerns. The goal is putting AI governance on the agenda.

Connect with unions. Even in non-union workplaces, unions can provide information, training, and support for organizing. The TUC, AFL-CIO, and sector-specific unions publish resources on AI and work.

Engage policy. Comment on proposed regulations, support legislation that protects workers, and vote for candidates who prioritize labor rights. Policy creates the conditions under which bargaining happens.

What Comes Next

The AI-labor moment is early. The Hollywood contracts are templates, not endpoints. European frameworks are being implemented, not yet fully tested. Warehouse organizing continues despite setbacks.

Several developments will shape the future:

Will organizing expand? If workers in tech, logistics, and other AI-intensive sectors build power, bargaining over AI will expand. If organizing stalls, protections will remain limited to unionized enclaves.

Will policy catch up? The EU is ahead, but implementation matters. U.S. federal policy remains largely absent, though state and local efforts continue. International coordination is minimal.

Will AI capabilities accelerate? Current bargaining addresses current AI. If AI capabilities leap forward — enabling displacement of occupations now considered safe — existing agreements may need renegotiation.

Will solidarity cross borders? AI deployment is global. Meaningful worker power requires international coordination. Some unions are exploring cross-border collaboration, but structures are nascent.

When AI enters the workplace, someone decides how it's used. Collective bargaining ensures that workers have a seat at that table. The question is whether enough workers will have access to that seat — and whether policy will extend the table to those currently excluded.

Redistribution through collective power. It's not automatic. It's fought for.

But the fight is producing results. The WGA and SAG-AFTRA contracts, European works council agreements, and even the early Amazon organizing demonstrate that workers can win meaningful protections — not by rejecting AI, but by insisting on a voice in how it is deployed. The most effective responses acknowledge that AI can genuinely improve work while demanding that the terms of that improvement are negotiated, not imposed. This is how technology has always been domesticated: not through resistance alone, but through the patient construction of institutions that balance power. The cooperative movement, democratic governance proposals, and sectoral bargaining models all represent paths toward that balance. The question is whether they will scale fast enough to matter.

Automation Debate
Manager Is Algorithm
Hidden Labor
AI Creative Industries
AI Coding Assistants
Freelance Markets
White-Collar Displacement
AI Cooperatives
Democratic AI Governance
EU AI Act Takes Effect
EU AI Act Explained
State-Level AI Regulation

Sources

Benanav, Aaron. Automation and the Future of Work. Verso, 2020. Publisher
Countouris, Nicola and Valerio De Stefano. "The EU Platform Work Directive and Platform Work." Comparative Labor Law & Policy Journal, 2024. SSRN
De Stefano, Valerio. "Negotiating the Algorithm: Automation, Artificial Intelligence, and Labor Protection." Comparative Labor Law & Policy Journal 41, no. 1 (2019). SSRN
IG Metall. "Artificial Intelligence in the Workplace: A Framework Agreement." 2022. IG Metall
Kellogg, Katherine C., Melissa A. Valentine, and Angèle Christin. "Algorithms at Work: The New Contested Terrain of Control." Academy of Management Annals 14, no. 1 (2020). AoM
Rosenblat, Alex. Uberland: How Algorithms Are Rewriting the Rules of Work. University of California Press, 2018. Publisher
SAG-AFTRA. "SAG-AFTRA Tentative Agreement Summary: Artificial Intelligence." November 2023. SAG-AFTRA
TUC. "Artificial Intelligence in the Workplace." 2021. TUC
TUC. "Technology Managing People: The Worker Experience." 2020. TUC
Writers Guild of America. "WGA 2023 MBA Summary: Artificial Intelligence." September 2023. WGA

General

Collective Bargaining in the Age of AI