In the summer of 2023, 116,000 writers and actors walked off their jobs simultaneously — the first Hollywood dual strike since 1960. At the center of their demands: binding limits on how studios could use AI to generate scripts, replicate performances, and scan background actors' likenesses. The strike lasted 148 days for writers, 118 days for actors. When contracts were finally signed, both the Writers Guild of America and SAG-AFTRA had secured explicit AI provisions — the first major AI clauses in American entertainment labor history.
Individual workers have almost no leverage when a company decides to automate a job. A single employee who objects to AI surveillance or an algorithmic scheduling system can be ignored or replaced. But a union representing thousands can extract binding commitments — and enforce them through grievance procedures, arbitration, or strike action.
The WGA and SAG-AFTRA agreements illustrate the range of possible AI provisions. The WGA deal required studios to disclose when AI-generated material is used and prohibited using AI to replace a minimum writing room. SAG-AFTRA's contract required explicit consent and additional compensation for digital replicas of a performer's face, voice, or likeness — with the consent requirement applying even to background extras whose image might be scanned and reused indefinitely.
These were not cosmetic wins. Studios had argued before the strike that AI scanning of extras required no separate payment and no ongoing consent. The union disagreed. The eventual contract language established the principle that workers own a property interest in their own likeness — a precedent with implications far beyond Hollywood.
The WGA agreement specified that AI-generated material cannot be used to undermine minimums, that writers cannot be required to write with AI tools, and that if a company uses AI to generate a first draft, a human writer hired to rewrite it receives full minimum compensation — the AI does not reduce the floor.
The Hollywood strike was the most visible, but not isolated. In 2022 and 2023, the International Brotherhood of Teamsters negotiated language at several logistics companies requiring advance notice before deploying autonomous warehouse equipment and guaranteeing retraining rights for displaced workers. UPS Teamsters in their 2023 contract secured provisions around surveillance technology — requiring disclosure of any new AI monitoring tools and giving workers a 90-day adjustment period before such tools could be used in discipline.
In Europe, IG Metall — the German metalworkers' union with roughly 2.2 million members — has been negotiating "works council" rights over algorithmic management systems since 2020. Under German co-determination law, works councils have legal standing to block or modify the introduction of technology that affects working conditions. IG Metall has used this leverage to require joint human-AI oversight committees at major automotive plants.
The Communications Workers of America launched a formal "AI Accountability" campaign in 2023, surveying members about AI tools already in use at their workplaces and drafting model contract language that locals can adapt. Their research found that over 60% of CWA members were already subject to at least one AI-driven monitoring or performance-evaluation system — most without any contract language governing it.
Labor researchers studying these negotiations have identified a common cluster of demands across sectors:
The WGA and SAG-AFTRA strikes demonstrated that AI is now a mainstream labor issue, not a futurist abstraction. The fact that 116,000 workers stopped production specifically over AI provisions — and won — signals to every industry that AI governance is bargainable. Workers in non-union sectors are watching these outcomes and using them as templates for workplace policy negotiations.
Union density in the United States has declined from roughly 35% of private-sector workers in the 1950s to under 6% today. The vast majority of American workers have no union and no collective bargaining. The WGA's wins apply only to WGA members — the millions of gig workers, warehouse pickers, and customer service agents using AI-mediated systems have no equivalent protection unless they organize or unless legislation fills the gap.
Even where unions exist, AI provisions face a structural challenge: technology changes faster than contract cycles. A provision negotiated in 2023 may not cover AI tools deployed in 2025. Some unions are responding by negotiating standing joint technology committees — ongoing bodies with authority to review new AI deployments between contract renewals, rather than waiting for the next bargaining round.
You represent workers at a large call center where the employer has just announced deployment of an AI system that will monitor every call in real time, score agent performance automatically, and flag underperforming employees for discipline — all without human review of the AI's decisions.
Use the AI assistant below to draft and refine contract language protecting your members. Ask it to help you write specific clauses, evaluate their strength, or compare them to real-world examples from WGA, SAG-AFTRA, or Teamsters agreements.
On January 1, 2023, New York City's Local Law 144 took effect — the first municipal law in the United States requiring employers to audit AI hiring tools for bias before using them and to publicly post the results. Any company using an automated employment decision tool to screen applicants or rank candidates in NYC must commission an independent bias audit annually and notify applicants that such a tool is being used. Violations carry fines of up to $1,500 per day. The law had been passed by the City Council in 2021 after a two-year lobbying campaign by worker and civil rights groups.
In the absence of comprehensive federal AI legislation, American workers face a patchwork of state and local rules. New York City's bias audit law is the most prominent, but it is narrow — it covers hiring tools, not performance monitoring, scheduling, or termination algorithms. Illinois passed the Artificial Intelligence Video Interview Act in 2019, requiring employers to explain how AI video interview analysis works before using it and to get written consent from applicants. The law also required employers to delete video data within 30 days of a request.
Maryland passed a similar law in 2020 covering facial recognition in job interviews. California's Consumer Privacy Act, while not AI-specific, gives workers and applicants some rights to know what data is collected about them and to opt out of its sale — rights that interact with AI systems in complex ways that are still being litigated.
Colorado passed a broad AI Act in 2024 requiring developers and deployers of "high-risk" AI systems — including employment decisions — to perform impact assessments and give affected individuals the right to appeal automated decisions. Colorado's law is the closest any U.S. state has come to the EU's approach.
The European Union's AI Act, finalized in 2024, takes a risk-based approach that creates binding obligations for AI systems used in employment, education, credit, and other consequential domains. AI tools used to recruit, screen, promote, or terminate workers are classified as "high-risk" — triggering requirements for human oversight, transparency documentation, bias testing, and worker notification.
Companies operating in the EU must maintain detailed technical documentation of high-risk AI systems, conduct conformity assessments, register systems in an EU database, and provide workers with meaningful information about how automated systems affect them. Non-compliance can result in fines of up to 3% of global annual turnover.
The EU AI Act is notable because it applies to any company using covered AI tools within the EU — including American multinationals. This means European workers at US-headquartered companies may have legal rights that their American counterparts at the same company lack. Labor researchers have described this as a potential "Brussels Effect" — European standards becoming de facto global standards because multinationals find it simpler to apply one compliance regime worldwide.
LL144 requires annual bias audits of automated employment decision tools used in hiring and promotion, public posting of audit results, and candidate notification. It does NOT cover AI used in performance monitoring, algorithmic scheduling, termination recommendations, or workplace surveillance. Civil society groups have already begun advocating for an expanded version covering these gaps.
Municipal action on AI has accelerated partly because cities are often the direct employers of workers most affected — transit workers, public hospital staff, city agency employees. Seattle passed a gig worker transparency ordinance in 2022 requiring app-based delivery companies to provide workers with detailed explanations of how pay is calculated, including any algorithmic components. The ordinance also required advance notice before algorithmic changes that would reduce earnings.
San Francisco's Board of Supervisors passed regulations in 2023 requiring city agencies to inventory AI tools in use and conduct public hearings before deploying any AI system that makes decisions affecting residents or city workers. The requirement applies to internal HR tools as well as public-facing systems.
These municipal actions create demonstration effects — when a city implements a requirement and enforcement doesn't collapse the economy, other cities gain confidence to follow. NYC's LL144 has already been cited as a model in proposed legislation in Chicago, Boston, and Philadelphia.
Legislation without enforcement is symbolic. NYC's LL144 illustrates the challenge: as of mid-2024, a relatively small number of bias audits had been publicly posted despite many employers clearly using covered tools. The city's enforcement agency was under-resourced, and some employers appeared to be waiting to see whether the law would actually be enforced before complying. Worker advocacy groups had to file complaints to prompt action, and legal challenges from employers had delayed implementation of key provisions.
This pattern — strong-sounding law, weak enforcement — is common in workplace regulation. The EU AI Act's larger fines and the threat of market exclusion may produce stronger compliance, but the regulation will not be fully enforced until the mid-2020s at the earliest.
Municipal and state-level AI laws matter because they establish legal frameworks, create enforcement precedents, and build public expectation that AI in employment must be transparent and auditable. Even imperfectly enforced, these laws shift the burden: employers must now justify their AI tools rather than simply deploying them. Workers and advocates who know these laws exist can use them as leverage even before formal enforcement catches up.
You work for a logistics company headquartered in New York City with operations in Illinois, Colorado, and Germany. Your employer has just rolled out three AI systems: (1) an automated video interview tool for new hires, (2) a real-time warehouse productivity monitoring system that automatically flags workers for discipline, and (3) an algorithmic scheduling system for delivery drivers.
Use the assistant below to identify which laws — NYC LL144, Illinois AI Video Interview Act, Colorado AI Act, EU AI Act — apply to each system, what rights workers have under each, and how to use those rights strategically.
In 2019, California passed AB5 — a law reclassifying most gig workers as employees, which would have entitled them to minimum wage, overtime, and the right to form unions. Uber and Lyft spent over $200 million on Proposition 22, a ballot initiative to exempt themselves from AB5. Prop 22 passed in November 2020 with 58% of the vote, enshrining a "third category" of worker in California law with some benefits but no collective bargaining rights. In 2021, a California Superior Court judge ruled Prop 22 unconstitutional. The legal battle continues. Meanwhile, the platforms maintained algorithmic control of their workforce throughout.
Platform companies have developed a distinctive model: use AI to exercise employer-like control over workers while maintaining their legal status as independent contractors. The algorithm sets prices, assigns work, monitors performance, deactivates accounts for policy violations, and adjusts earnings dynamically — all functions traditionally performed by employers — but without the legal obligations of employment.
Research by sociologist Alex Rosenblat, published in her 2018 book Uberland, documented how Uber's app creates what she called "algorithmic management" — a system of incentives, information asymmetries, and automated enforcement that shapes driver behavior without any human manager giving direct instructions. Drivers who don't follow the algorithm's implicit guidance — by refusing surge pricing areas, maintaining low acceptance rates, or logging on during unprofitable periods — see their earnings fall or their accounts restricted.
This model creates an enforcement gap: traditional labor law governs the employment relationship, but platform workers technically have no employer. They cannot collectively bargain under the National Labor Relations Act, which covers employees, not independent contractors.
Without NLRA protections, platform workers have developed alternative organizing strategies. In 2019, Rideshare Drivers United — a California-based worker organization — coordinated a driver strike on May 8, 2019, the day before Uber's IPO, timed specifically for maximum media and investor attention. Thousands of drivers in Los Angeles, San Francisco, and other cities logged off the app for several hours. The action generated extensive coverage and established that coordinated platform worker action was possible without formal union status.
In New York, the Independent Drivers Guild — which represents Uber and Lyft drivers but is technically not a union under federal law — negotiated directly with the platforms to establish minimum earnings floors after NYC's Taxi and Limousine Commission set new per-mile rates in 2019. The IDG used regulatory engagement rather than collective bargaining to achieve wage increases — working with city government to impose rules the platforms could not avoid.
Internationally, Deliveroo riders in the UK organized through the Independent Workers' Union of Great Britain, which won a legal ruling in 2021 that Deliveroo riders could collectively bargain — a decision that remains contested but established a precedent. In Spain, the "Riders' Law" (Royal Decree-Law 9/2021) required all food delivery platforms to reclassify delivery workers as employees, a direct legislative response to platform worker organizing.
Multiple platform worker organizations have converged on a common demand: algorithmic transparency. Workers want to know the specific criteria by which their accounts are rated, deactivated, or de-prioritized. This demand reframes the issue from "wages" to "due process" — asking for the same right any employee would have to understand why they were disciplined. Transparency demands have succeeded in extracting policy disclosures from several platforms that previously provided none.
Amazon workers are not gig workers, but their organizing confronts algorithmic management intensively. When workers at the Staten Island JFK8 warehouse voted to form the Amazon Labor Union in April 2022 — the first successful union election at an Amazon facility in the US — one of the central organizing issues was Amazon's Time Off Task (TOT) system, an AI-driven monitoring tool that tracked workers' movements and could automatically trigger discipline if workers spent too much time away from their scanning stations.
Workers described the TOT system as creating conditions in which they felt unable to take bathroom breaks. Organizers used this grievance as a central recruiting message. The vote — 2,654 for the union versus 2,131 against — reflected a workforce that had concluded algorithmic management had crossed a line that required collective response. Subsequent ALU organizing efforts at other Amazon facilities, while less successful, consistently cited algorithmic monitoring as a core issue.
Platform workers have used the same digital infrastructure that employs them to organize against it. Rideshare drivers communicate through Facebook groups, WhatsApp chains, and Telegram channels that coordinate actions, share information about algorithm changes, and mobilize responses to policy changes. In 2021, Instacart shoppers used an app — built by workers themselves — to identify and flag low-pay batches in real time, allowing collective action to leave poorly paid orders unclaimed.
This "counter-algorithmic" organizing — using technology to respond to technology — represents a genuinely new form of labor action. Workers cannot formally strike, but they can collectively decline unprofitable work, share information asymmetries that the algorithm exploits, and coordinate public pressure campaigns.
Platform workers have developed a repertoire of collective action tools that operate outside traditional labor law — regulatory engagement, strategic timing of actions, digital coordination, and algorithmic transparency demands. These approaches are less powerful than formal collective bargaining but have produced real gains in specific contexts. The crucial variable is whether a regulatory body (city, state, EU) can be persuaded to impose the standards the platforms will not negotiate voluntarily.
You're an organizer working with food delivery riders in a major U.S. city. Your riders are classified as independent contractors, so the NLRA doesn't apply. The platform recently changed its algorithm, cutting base pay per delivery by 12% without notice. Riders want to respond collectively but don't know what options they have.
Use the assistant below to design a multi-pronged strategy: identify which approaches (regulatory engagement, coordinated work refusal, media pressure, legal challenges, digital organizing) apply in this situation, analyze their risks and likely effectiveness, and develop a realistic action plan.
When Carnegie Mellon University and Uber began developing autonomous vehicle technology in Pittsburgh, the city had a choice: allow the wealth and disruption to flow through without community engagement, or negotiate. Pittsburgh's Mayor Bill Peduto pursued a community benefit framework — working with neighborhood groups, labor organizations, and CMU to ensure that the jobs created by AV research were accessible to local residents, that testing did not disproportionately burden low-income neighborhoods, and that workforce development programs were integrated into the research ecosystem. The effort was incomplete and contested, but it established that a community's consent to be a technology laboratory was negotiable.
Worker centers are community-based organizations that serve low-wage workers who lack union representation — domestic workers, day laborers, restaurant workers, home health aides, and increasingly gig workers. Unlike unions, they are not governed by the NLRA and can be more flexible in their organizing and advocacy approaches. As of 2023, there are over 200 worker centers in the United States, many affiliated with the National Day Laborer Organizing Network or the National Domestic Workers Alliance.
Worker centers have begun building specific AI literacy and advocacy capacity. The Restaurant Opportunities Centers United launched programs in 2023 helping restaurant workers understand AI scheduling tools — systems that optimize labor costs by calling workers in on short notice or sending them home early based on algorithmic demand forecasts. ROC's approach combined legal education (what scheduling laws apply) with organizing support (how to raise the issue with employers collectively) and policy advocacy (pushing for predictive scheduling laws in more cities).
Community benefit agreements (CBAs) are legally binding contracts between developers or employers and community coalitions, negotiated as a condition of receiving public subsidy, zoning approval, or other government benefits. They have been used for decades around real estate development — requiring local hiring, affordable housing, and living wages. They are now being adapted to technology deployments.
In 2023, when Kansas City expanded its Smart City infrastructure — including AI-enabled traffic and public safety systems — community organizations negotiated a formal data governance agreement specifying that data collected in predominantly Black neighborhoods would not be shared with federal immigration authorities and that algorithmic policing tools would require city council approval before deployment. The agreement did not stop the technology rollout, but it created a governance framework and enforcement mechanism that residents could invoke.
Amazon's decision to locate HQ2 in Northern Virginia and New York City (partially, before NYC's portion was cancelled) generated extensive CBA negotiation discussions. When the NYC deal fell apart in 2019 partly over community opposition, one factor was Amazon's refusal to accept any binding commitment on union rights or local hiring. The episode illustrated both the potential and limits of community leverage over major technology employers.
The NDWA, which represents approximately 2.2 million domestic workers including nannies, house cleaners, and home health aides, launched a formal AI advocacy program in 2023. Their central concern: AI tools being used by staffing agencies and care platforms to screen, rank, and deactivate domestic workers based on algorithmic scores that workers cannot see or contest. The NDWA has pushed for a "Domestic Workers' Bill of Rights for the AI Era" that would extend algorithmic transparency and human review requirements specifically to care work platforms.
Several regional economies have established multi-stakeholder coalitions to address AI-driven workforce transitions. The Greater Washington Partnership — a coalition of major employers, universities, and workforce organizations in the DC-Baltimore corridor — launched a "Capital CoLAB" initiative specifically designed to build AI and data science talent pipelines in communities underrepresented in the technology sector. The initiative combined employer commitments to hire graduates with community college curriculum development and paid apprenticeships.
In the Midwest, the Chicago Cook Workforce Partnership has worked with local employers to map which occupations in the Chicago metro area are most exposed to AI displacement and to pre-position retraining resources in those communities before displacement peaks rather than after. Their 2023 workforce outlook specifically identified clerical, customer service, and logistics roles as requiring transition support within 3–5 years and began building partnerships with community colleges to develop relevant credentials.
These regional coalitions differ from individual employer retraining programs in a crucial way: they are designed to serve workers across employers, which means workers who are displaced from one company can access support without depending on that company's goodwill or solvency.
Community colleges are positioned as the most accessible retraining infrastructure for workers displaced by AI — they are geographically distributed, affordable, and have established relationships with local employers. But their capacity to adapt quickly is limited. Curriculum development cycles can take 2–3 years, and faculty hiring in technical fields competes with private sector salaries the colleges cannot match.
Several states have experimented with faster pathways. Registered Apprenticeship programs — which combine on-the-job training with classroom instruction, are employer-funded, and result in portable credentials — have been expanded beyond traditional trades into technology fields. The Biden administration's 2021 executive order expanding apprenticeship programs in AI and cybersecurity created a framework that community organizations and unions have since used to establish new programs.
The fundamental challenge remains funding: retraining is expensive, the workers who need it most have the least capacity to pay for it, and employer funding is voluntary and inconsistent. The most successful regional programs have combined employer contributions, state funding, and federal workforce development grants — a coalition of funding sources that reflects the coalition of institutions needed for adaptation at scale.
Collective adaptation to AI is not just about unions or legislation — it requires building durable community institutions: worker centers that provide AI literacy and advocacy capacity for unorganized workers; CBAs that embed governance conditions into technology deployments; regional retraining coalitions that serve workers across employers; and public institutions with sustainable funding. The communities that navigate the AI transition most successfully will be those that invested in this infrastructure before displacement peaked, not after.
You're working with a coalition in a mid-sized Midwestern city where the largest employer — a regional bank with 4,000 workers — has just announced it will deploy AI to handle customer service, back-office processing, and loan underwriting. The company estimates 1,200 jobs will be eliminated over 3 years. The workforce is predominantly women of color; the city's unemployment rate is already 6.2%.
Use the assistant to design a comprehensive community response: identify which institutions (worker center, community college, city government, CBAs, regional coalition) should play what roles, what a realistic timeline looks like, and what the first 90-day priorities should be.