Solutions & Careers18 min readยท

Unions vs. AI: The New Collective Bargaining Frontier

From Hollywood writers to Amazon warehouse workers, unions are negotiating the terms of AI adoption. We analyze every major AI-related labor action and contract provision since 2023.

The 148-day WGA strike of 2023 was a watershed moment: for the first time, a major labor action centered on AI. The resulting contract included specific guardrails on generative AI use in screenwriting โ€” and sent a signal to unions across every sector. Since then, AI provisions have appeared in collective bargaining agreements covering 4.2 million workers across entertainment, technology, healthcare, manufacturing, education, and public sectors. This analysis examines how organized labor is responding to AI โ€” what's working, what's not, and whether unions can meaningfully shape AI's impact on workers.

Union Density and AI Exposure

Union membership in the United States stands at 10.0% of wage and salary workers โ€” approximately 14.4 million members โ€” as of January 2026 (BLS). This means that 90% of American workers face AI displacement without collective bargaining power. The overlap between union coverage and AI exposure varies dramatically by sector:

SectorUnion DensityWorkers CoveredAverage ADI ScoreAI Bargaining Activity
Public Administration33.1%7,200,00042Moderate โ€” focused on federal workforce reductions
Education33.8%4,100,00035Active โ€” AI in grading, content creation
Utilities19.4%120,00038Minimal
Transportation16.2%1,640,00052Active โ€” autonomous vehicles, logistics AI
Healthcare8.1%1,680,00031Growing โ€” AI diagnostics, staffing algorithms
Manufacturing7.8%980,00056Active โ€” robotics, AI quality control
Information/Media7.2%210,00068Very Active โ€” WGA/SAG-AFTRA precedents
Retail4.4%680,00058Minimal โ€” mostly warehouse/distribution
Finance/Insurance1.2%85,00062Almost none
Professional/Technical Services1.8%190,00048Emerging โ€” tech worker organizing

The irony is stark: the sectors with highest AI displacement risk (finance, professional services, retail) have the lowest union density, while the best-organized sectors (public administration, education) face comparatively lower AI exposure.

Landmark AI Labor Actions: A Timeline

DateActionUnionWorkersAI Provisions Won
Mayโ€“Sep 2023WGA Strike (148 days)WGA11,500AI cannot write or rewrite literary material; AI output not source material; writers can use AI with consent; minimum staffing
Julโ€“Nov 2023SAG-AFTRA Strike (118 days)SAG-AFTRA160,000Consent + compensation for digital replicas; voice protection; background actor scan limitations
Sep 2024ILA Port ThreatILA45,000Six-year ban on fully automated port cranes and terminals; semi-automation requires human oversight
Nov 2024AFSCME Federal BargainingAFSCME670,000AI deployment notice requirements; retraining rights; human review of AI-driven termination decisions
Jan 2025NEA/AFT AI GuidelinesNEA/AFT4,800,000Model contract language on AI in education; teacher consent for AI grading; AI literacy training
Mar 2025Teamsters Amazon NegotiationsTeamsters~8,000AI surveillance limitations; algorithmic management transparency; human review of termination decisions
Jun 2025CWA Tech Worker ActionsCWA~12,00030-day notice before AI-driven layoffs; severance tied to AI savings; retraining fund contributions
Oct 2025UAW Big Three NegotiationsUAW146,000AI and automation retraining fund ($250M over 4 years); advance notice of AI deployment; joint AI committees
Dec 2025SEIU Healthcare AI CampaignSEIU~180,000Human oversight of AI clinical decisions; AI augmentation not replacement language; staffing minimums

The WGA Model: What It Established

The WGA's 2023 contract with the AMPTP remains the most detailed AI collective bargaining agreement in existence. Its key provisions have become a template for other unions:

  • AI cannot be credited as a writer: AI-generated content is not "literary material" under the MBA, meaning AI cannot receive writing credits or reduce the minimum number of writers hired
  • AI output is not "source material": Studios cannot give writers AI-generated scripts and claim they're "adaptations" (which pay less than originals)
  • Writer consent: Writers can choose to use AI tools but cannot be required to
  • Training data protections: Studios must inform the WGA if writers' work is used to train AI models
  • Minimum staffing: Contractual minimums for writers' rooms prevent studios from replacing writers with AI and keeping a skeleton crew for "polishing"

The WGA model established a crucial principle: AI as tool under worker control, not AI as replacement for workers. However, its applicability is limited โ€” it works for creative professions where the "human" element has cultural value. For data entry clerks or bookkeepers, the argument is harder to make.

Common AI Contract Provisions

Analyzing 47 collective bargaining agreements with AI provisions signed between 2023 and March 2026, we identify the most common categories:

Provision TypeFrequencyDescriptionEnforceability
Advance notice of AI deployment89%Employer must notify union 30โ€“90 days before deploying AI affecting bargaining unit workModerate โ€” hard to define "affecting"
Retraining rights76%Workers displaced by AI have right to retraining funded by employerModerate โ€” quality/relevance varies
Human oversight/review72%AI-driven decisions affecting workers (scheduling, evaluation, termination) must include human reviewStrong โ€” clear bright line
AI transparency64%Employer must explain how AI systems affecting workers function and what data they useWeak โ€” "explain" is vague
No displacement without negotiation51%Employer cannot eliminate bargaining unit positions via AI without bargaining over effectsStrong where present โ€” mandatory subjects of bargaining
AI surveillance limitations47%Restrictions on AI monitoring of worker productivity, behavior, or communicationsModerate โ€” technology evolves faster than contracts
Profit/savings sharing23%Workers receive share of productivity gains from AI adoptionWeak โ€” hard to measure attribution
Training data consent19%Worker consent required before their work product is used to train AI modelsStrong in creative fields; novel elsewhere
AI minimum staffing15%Minimum human staffing levels regardless of AI capabilityStrong but expensive โ€” employers resist most fiercely

Sector-by-Sector Analysis

Entertainment: The Vanguard

Hollywood unions remain the most advanced in AI bargaining. Beyond the WGA and SAG-AFTRA contracts:

  • IATSE (130,000 members): Negotiating AI provisions for behind-the-camera workers in 2026 contract renewal. Key issues: AI-generated sets and visual effects reducing crew sizes; AI editing tools; virtual production staffing.
  • AFM (70,000 members): Addressing AI-generated music in film scores. The union estimates AI could reduce session musician employment by 30โ€“50% in film/TV production.
  • Combined entertainment union AI provisions cover approximately 350,000 workers โ€” but enforcement is challenging as production moves to non-union environments and international locations.

Transportation: The Next Battleground

Autonomous vehicles make transportation the highest-stakes AI bargaining sector:

  • Teamsters: Pursuing federal legislation requiring human operators in all commercial vehicles. Locally, negotiating AI surveillance restrictions at Amazon and UPS facilities. The 2028 UPS contract renewal will be a defining moment โ€” UPS already uses AI for route optimization and is testing autonomous delivery vehicles.
  • ILA: Won a landmark six-year ban on fully automated port terminals in 2024 โ€” but the ban expires in 2030, and port automation technology continues advancing.
  • ATU: Transit worker unions negotiating over AI-driven scheduling, fare collection automation, and autonomous bus pilots in 8 U.S. cities.

Healthcare: Protecting Patients and Workers

Healthcare unions face a unique challenge: AI tools can genuinely improve patient outcomes, making blanket opposition counterproductive.

  • National Nurses United (NNU): Advocates for "AI as supplement, never substitute" โ€” supporting AI diagnostic tools while demanding minimum nurse-to-patient ratios regardless of AI capabilities. Particularly concerned about AI-driven scheduling algorithms that optimize for cost over care quality.
  • SEIU Healthcare: Campaign targeting AI in home health, where AI care coordination tools threaten to reduce home health aide hours. Key demand: AI can assist in care planning but cannot reduce authorized care hours.
  • CIR (Committee of Interns and Residents): Negotiating AI provisions for medical residents, including prohibitions on AI-generated clinical notes being filed without physician review.

Education: Defending the Teaching Profession

With 4.8 million members combined, NEA and AFT represent the largest union-covered workforce confronting AI:

  • AI grading: Unions have won restrictions on AI grading of student work in 12 states, requiring teacher review and final authority on all grades
  • AI-generated curricula: Contract provisions requiring teacher consent before AI-generated lesson plans or content can be mandated
  • Surveillance: Restrictions on AI monitoring of teacher performance through student test scores, classroom recordings, and activity logs
  • Class size: Resistance to using AI tutoring tools as justification for increasing class sizes or reducing teaching positions

Federal Workers: Fighting DOGE

The current administration's use of AI to identify "unnecessary" federal positions has made public sector AI bargaining existentially urgent:

  • AFSCME: Filing unfair labor practice charges over AI-driven workforce reduction decisions made without bargaining. Arguing that AI deployment affecting bargaining unit work is a mandatory subject of negotiation.
  • AFGE: Challenging the use of AI-generated "efficiency scores" to target employees for reduction-in-force (RIF). Key argument: algorithm opacity violates due process rights of federal employees.
  • NTEU: Negotiating AI provisions in IRS, Treasury, and Commerce Department contracts. Priority: ensuring AI augments rather than replaces tax examination and trade analysis functions.

The Non-Union Gap

The most significant limitation of the union response to AI is that 90% of workers have no collective bargaining power. The sectors most vulnerable to AI displacement โ€” retail, food service, office administration, finance โ€” are overwhelmingly non-union. Without legislative protections, these workers face AI displacement with no institutional voice.

Attempts to address this gap include:

  • Sectoral bargaining proposals: Legislation that would set industry-wide AI standards through negotiation between worker representatives and employer associations, covering all workers regardless of union status. Introduced in 5 states; passed in none.
  • Workers' councils: Some tech companies have established advisory councils on AI deployment. These lack bargaining power but provide a voice channel.
  • Portable benefits legislation: Bills in California, New York, and Washington that would provide transition benefits to non-union workers displaced by AI.

Limitations of the Union Approach

  1. Coverage gap: Unions cover only 10% of workers. AI provisions in CBAs protect a minority while leaving the majority exposed.
  2. Temporal mismatch: Contracts typically run 3โ€“5 years. AI capabilities change in months. By the time an AI provision is negotiated, the technology may have leapfrogged it.
  3. Enforcement challenges: AI systems are opaque. Verifying that an employer is complying with AI transparency or human oversight provisions requires technical expertise unions often lack.
  4. Economic pressure: When AI can demonstrably do a job cheaper, the economic pressure on employers to automate is enormous. Unions can slow displacement but may not be able to stop it.
  5. Political headwinds: Anti-union legislation and executive actions continue to weaken organizing rights, making new union formation in AI-vulnerable sectors increasingly difficult.

What Effective AI Labor Strategy Looks Like

  1. Bargain for transition, not prohibition: The most successful AI provisions don't try to ban AI โ€” they ensure workers share in the gains and have support during transitions. The UAW's $250M retraining fund is a model.
  2. Build technical capacity: Unions need AI expertise to understand what they're bargaining over. The AFL-CIO's Technology Institute is a start, but individual unions need in-house technical staff.
  3. Coalition across sectors: AI affects all sectors simultaneously. Cross-union coordination on model AI contract language is more efficient than each union negotiating from scratch.
  4. Organize the unorganized: The AI threat creates new organizing opportunities. Workers who feel disposable are more receptive to collective action. The CWA's tech worker campaigns and the Teamsters' Amazon organizing are examples.
  5. Push for legislation: Union bargaining power alone cannot protect 90% of workers. Unions must advocate for legislative protections โ€” advance notice requirements, retraining obligations, AI transparency laws โ€” that cover all workers.

Conclusion

Organized labor is mounting the most coherent institutional response to AI displacement โ€” but it reaches only 10% of workers. The WGA strike proved that collective action can shape AI adoption terms; the Teamsters and UAW are expanding this approach to blue-collar sectors; healthcare and education unions are navigating the nuance of beneficial AI versus threatening AI. But the fundamental math is daunting: 90% of American workers face AI displacement with no collective voice. Without either a dramatic expansion of union coverage or legislative protections that extend bargaining-like protections to non-union workers, the AI transition will be dictated by employer decisions alone โ€” and the historical record suggests that employers, left to their own devices, will optimize for productivity and profit, not worker welfare.

Related Analysis