Amazon warehouse workers in Coventry walked off the job in January 2023, becoming the first Amazon UK employees to officially strike. Their grievances were not abstract. The "rate" — a productivity metric set by algorithmic monitoring systems — governed every minute of their shifts. Miss the rate too many times and the system flagged you for discipline, sometimes terminating workers without direct human review.
Workers described the same dynamic seen in Amazon fulfillment centers globally: an AI-mediated management layer that counted steps, measured idle seconds, and generated performance reports faster than any human supervisor ever could. The strike lasted weeks and drew international attention to a question that labor law had not yet clearly answered.
The Amazon Coventry strike was not an isolated flash point. It crystallized a structural shift in how large employers manage labor: the replacement of supervisor discretion with algorithmic directive. Understanding this shift is essential to understanding contemporary labor relations.
At Amazon warehouses globally, the Time Off Task (TOT) system automatically tracks seconds workers spend not scanning items. Internal documents revealed in a 2021 investigation by The Verge showed that TOT data fed directly into termination decisions — with algorithms issuing warnings and eventually firing workers without a manager signing off. Amazon confirmed the system existed but stated human review occurred before final termination.
The practice is not limited to warehouses. In 2022, the App-Based Drivers Association in the UK won a Supreme Court ruling that Uber drivers were workers entitled to minimum wage and holiday pay — a decision partly enabled by documenting how Uber's dispatch algorithm controlled drivers' economic conditions so thoroughly that the platform functioned as an employer regardless of its contractual framing.
A 2021 investigation by journalists at The Guardian and The Verge found that Amazon's internal warehouse management system had automatically terminated hundreds of workers in the US for failing algorithmic productivity thresholds, often with no direct human decision-maker involved. Amazon later said it had modified the system to require human review.
Algorithmic management is most visible in gig-economy platforms but has spread into traditional employment relationships. The key mechanisms are: automated performance scoring, real-time behavioral monitoring, dynamic wage-setting, and opaque deactivation.
Uber's "surge pricing" algorithm sets wages unilaterally and in real time. Deliveroo's "Frank" algorithm — named and discussed in leaked internal documents — calculated driver pay based on speed and distance models invisible to drivers. Instacart's batch-assignment algorithm determines which shoppers receive the most profitable orders. In each case, the algorithm functions as a de facto manager, setting pace, pay, and conditions without the legal accountability traditionally attached to employers.
California's Proposition 22 (2020) and subsequent legal battles across Europe have forced courts and regulators to grapple with exactly this question: does algorithmic control constitute employment? The answer has enormous consequences for collective bargaining rights, because you generally cannot unionize against a platform if you are legally classified as an independent contractor.
Traditional collective bargaining assumes identifiable management counterparts who exercise discretion over wages and working conditions. Algorithmic management disrupts this model in three ways.
First, it disaggregates accountability: when a worker is disciplined, the system issued the warning, not a named manager. Second, it generates information asymmetry: management has granular data on each worker's output; workers often cannot see the full logic of the system rating them. Third, it accelerates the pace of management decisions faster than grievance procedures were designed to handle.
The 2022–2023 UPS contract negotiations surfaced this explicitly. The Teamsters union negotiated specific language around surveillance technology and algorithmic monitoring, requiring the company to disclose new monitoring tools and bargain over their implementation. This represented one of the first major US collective bargaining agreements to address AI management tools directly.
Labor law frameworks built in the 20th century assume that management decisions flow from humans with identifiable authority. Algorithmic management creates a gap between legal accountability and operational control — a gap that unions, regulators, and courts are still working to close.
You will discuss documented cases of algorithmic management with an AI assistant trained on this module's content. Explore the legal, ethical, and labor-relations implications of systems like Amazon's TOT tracker and Uber's dispatch algorithm.
For the first time since 1960, both the Writers Guild of America (WGA) and the Screen Actors Guild–American Federation of Television and Radio Artists (SAG-AFTRA) were on strike simultaneously. The combined work stoppage, which ran from May through November 2023, effectively shut down Hollywood production and produced the most consequential entertainment labor agreements in decades.
AI was not the only issue — residuals and streaming economics were central — but the AI provisions attracted the most urgent attention. Studios had been exploring using generative AI to write script variations, create background characters, and — in some documented cases — scanning actors' likenesses for use without additional compensation.
The WGA's position was grounded in specific practices already underway. Studios had begun asking writers to "punch up" AI-generated scripts — meaning the human writer would revise material initially produced by a large language model — at rates below standard script fees. The Alliance of Motion Picture and Television Producers (AMPTP) initially resisted any restrictions on using AI as a writing tool.
On the actors' side, the issue was more visceral. SAG-AFTRA documented cases where background actors had been paid a flat daily rate to have their full-body digital scan captured, with contracts that studios argued gave them perpetual rights to use those likenesses in any production — effectively purchasing a digital replica of a person for a one-time fee. The union called this a "digital replica" threat and made it a central demand.
The studios also explored AI voice cloning. Audiobook narrators — many of them SAG-AFTRA members — found their recorded performances being used to train text-to-speech models. Some discovered their voices being sold as commercial products without consent or payment.
During the 2023 SAG-AFTRA strike, union representatives presented evidence that one major studio had scanned the likenesses of background performers under a contract that, the studio claimed, granted the right to use those digital replicas in perpetuity across any project. This specific practice became a flashpoint in negotiations and was directly addressed in the final agreement.
The WGA deal, ratified in September 2023, established landmark AI provisions. Studios cannot use AI to write or rewrite literary material, and AI-generated material cannot be used to undermine the minimum compensation standards for writers. Crucially, if a writer uses AI tools with company approval, the company cannot claim that reduces the writer's credit or compensation. The agreement also required studios to disclose when they provide AI-generated material to writers.
The SAG-AFTRA agreement, reached in November 2023, addressed digital replicas directly. Performers must provide informed consent for any AI replica of their likeness or voice, and separate negotiated compensation is required for each specific use — not a blanket perpetual license. The agreement also prohibited the use of AI to replace background actors at rates below scale and established a "digital double" framework requiring individual consent and payment.
Both agreements included provisions requiring ongoing joint AI committees — labor and management working together to address new AI applications as technology evolves, acknowledging that the technology would move faster than any single contract could anticipate.
The entertainment industry agreements established several contractual precedents that labor advocates in other industries immediately cited. The concept of informed consent before AI replication translated directly into discussions about AI voice systems in call centers, AI-generated likenesses in advertising, and AI tools in journalism and publishing.
The News Media Guild, representing journalists at outlets including The New York Times, began negotiating AI provisions in 2023 and 2024, explicitly referencing the Hollywood model. The Times had already filed a lawsuit against OpenAI and Microsoft in December 2023 over the use of its journalism to train AI models — an action that ran parallel to, but was legally distinct from, labor negotiations.
The broader precedent was structural: AI must be a subject of mandatory bargaining, not a unilateral management decision. Whether courts and labor boards will enforce this interpretation broadly remains an evolving question.
The 2023 Hollywood strikes produced the first detailed contractual language governing generative AI in a major industry. They demonstrated that organized labor can extract binding AI governance provisions through collective action — and that the specific demands (consent, compensation, disclosure, joint oversight) offer a template for other sectors to follow or adapt.
In this lab you will work through the practical challenge of translating the principles from the WGA and SAG-AFTRA agreements into bargaining language for other industries. The AI assistant will help you analyze existing contract provisions and think through how to adapt them.
When the European Commission published its proposed AI Act in April 2021, employment AI occupied a specific category: high risk. Systems used for recruitment, performance evaluation, task allocation, termination, and monitoring of workers were explicitly listed as requiring conformity assessments, transparency obligations, and human oversight mechanisms. This was not accidental. European labor federations had been lobbying intensively for exactly this classification, arguing that employment contexts warranted heightened protection regardless of the AI's function.
The EU AI Act, finalized in 2024 after three years of negotiation, established a risk-based framework. AI systems classified as high-risk in employment contexts — including CV screening, automated performance scoring, behavioral monitoring tools, and systems that make or significantly influence termination decisions — face the most stringent requirements.
These requirements include: maintaining detailed technical documentation; conducting conformity assessments before deployment; ensuring adequate human oversight; providing transparency to affected workers about the use of AI; and logging system decisions for audit purposes. Employers in the EU must register high-risk employment AI systems in a public database managed by the European AI Office.
Worker representatives are granted specific rights: the right to be informed about AI systems used in their workplace and the right to request explanations of significant decisions affecting their employment. For unionized workers, this creates new information rights that can feed directly into collective bargaining.
Article 26 of the EU AI Act imposes obligations on deployers of high-risk AI systems, including requiring that workers subject to AI monitoring or evaluation be informed of the AI system's use. Employers who deploy high-risk employment AI without conducting required conformity assessments face fines of up to €15 million or 3% of global annual turnover.
The United States has not passed comprehensive federal AI legislation as of 2024, but a patchwork of regulatory actions, state laws, and agency guidance creates a complex landscape. The Equal Employment Opportunity Commission (EEOC) issued guidance in 2023 clarifying that employers using AI in hiring and performance management remain liable for discriminatory outcomes under Title VII, the ADA, and the ADEA — even if the discrimination was introduced by an algorithm they purchased from a vendor.
Illinois passed the Artificial Intelligence Video Interview Act in 2019, requiring employers to notify candidates when AI is used to analyze video interviews, explain how the AI works, and obtain consent. New York City's Local Law 144, effective since July 2023, requires employers and employment agencies using automated employment decision tools to conduct annual bias audits and publish the results publicly.
The National Labor Relations Board (NLRB) has increasingly weighed in. General Counsel Jennifer Abruzzo issued a memo in October 2022 arguing that employers' use of surveillance technologies — including AI monitoring tools — could constitute unlawful interference with workers' Section 7 rights to organize and engage in concerted activity if the surveillance chills protected activity.
The United Kingdom, post-Brexit, has taken a different path from the EU. Rather than a comprehensive AI Act, the UK published an AI Regulatory Framework in 2023 that relies primarily on existing sector-specific regulators — the Information Commissioner's Office (ICO) for data protection, the Equality and Human Rights Commission for discrimination, and the Health and Safety Executive for workplace safety.
The ICO's Employment Practices and Data Protection code provides the most directly relevant guidance: employers using AI for monitoring or performance management must conduct data protection impact assessments (DPIAs), provide clear privacy notices to workers, and ensure that significant decisions affecting workers are not made solely by automated means without human review — a requirement derived from UK GDPR Article 22.
The 2023 Uber Supreme Court decision, and subsequent cases involving Deliveroo and Amazon, have also developed employment status case law that effectively holds that algorithmic control over working conditions can constitute an employment relationship regardless of how contracts are drafted — a precedent with significant implications for platform workers' organizing rights.
The EU has the most comprehensive mandatory framework; the US relies on a patchwork of agency guidance, state laws, and civil rights statutes; the UK is adapting existing regulatory structures. In all three jurisdictions, the direction of travel is toward greater transparency and accountability for employment AI — but enforcement capacity and speed of legal development lag far behind the pace of technology deployment.
You are advising a mid-sized logistics company operating in both the EU and the US that wants to deploy an AI system to monitor driver performance and flag unsafe behavior. Work through the regulatory compliance questions with the AI advisor.
Before the AI governance conversation had fully reached North America, European works councils were already developing tools for it. At Deutsche Telekom, the German works council (Betriebsrat) exercised its co-determination rights under the Works Constitution Act to negotiate binding agreements on the use of algorithmic management software. The company could not deploy new monitoring tools without the works council's agreement — a statutory right that existed independently of any collective bargaining agreement and that made the German model the most frequently cited international reference point for worker AI governance.
Germany's Works Constitution Act (Betriebsverfassungsgesetz) gives works councils the right to co-determine the introduction of technical systems that monitor worker behavior or performance. This right predates AI but applies fully to algorithmic management tools. Before an employer can deploy performance-monitoring software, task-allocation algorithms, or behavioral scoring systems, the works council must consent — or a labor court must rule on the dispute.
This framework has produced detailed works agreements (Betriebsvereinbarungen) at major German employers. Volkswagen, Siemens, Deutsche Bank, and SAP have all negotiated binding AI governance agreements with their works councils specifying what data AI systems may collect, how long it is retained, who may access it, what decisions it may influence, and what appeal mechanisms workers have. These agreements are often more detailed and protective than any national law requires.
The EU's European Works Council Directive and the Platform Work Directive (finalized in 2024) extend elements of this model across the EU, requiring meaningful consultation with worker representatives before deploying significant AI management systems in covered organizations.
Germany's largest industrial union, IG Metall, established an AI observatory in 2018 to document algorithmic management practices across manufacturing and collect data from works councils on what AI tools employers were deploying. By 2023, the observatory had documented over 200 cases and produced standardized contract language that works councils across Germany could adapt — a knowledge-sharing infrastructure that gave workers more negotiating capacity than they would have had individually.
US unions, operating without statutory co-determination rights, have relied on a combination of collective bargaining, regulatory advocacy, and direct organizing to address employment AI. The strategies that have gained traction fall into four categories.
Transparency demands: Requiring employers to disclose what AI systems they use, what data those systems collect, and what decisions they influence — before deployment, not after. The Communications Workers of America (CWA) made this a central demand in technology sector negotiations, and the 2023 contracts at several cable and telecommunications companies included disclosure language.
Joint oversight structures: Establishing labor-management committees with actual authority to review and approve new AI deployments, modeled on the Hollywood joint committees. The United Auto Workers (UAW) included AI governance provisions in its 2023 contract demands against the Detroit Three automakers, specifically requesting joint oversight of automation and AI tools in production settings.
Impact bargaining: Under the NLRA, employers generally must bargain with unions over the impacts and effects of management decisions even when the decisions themselves are not mandatory subjects of bargaining. Unions have argued that when AI tools significantly change working conditions, the employer must bargain over the impacts — even if the decision to use AI is itself a management prerogative. The NLRB's 2022 and 2023 guidance has supported this interpretation.
Organizing around AI grievances: The Amazon Labor Union's victory at the Staten Island JFK8 warehouse in April 2022 — the first successful Amazon union election in the US — was driven in significant part by worker frustration with algorithmic management, productivity quotas, and TOT monitoring. Organizers explicitly named the algorithm as a workplace hazard during the campaign.
Drawing on the cases and frameworks in this module, effective worker-protective AI governance in labor relations combines several elements. The EU model contributes mandatory risk classification and conformity requirements. The Hollywood model contributes consent, compensation, and joint oversight for creative and identity-based AI use. The German model contributes statutory co-determination and detailed works agreements. The US model contributes anti-discrimination enforcement, impact bargaining rights, and the organizing energy that comes from workers experiencing AI management directly.
The International Labour Organization (ILO) published a comprehensive framework paper in 2023 identifying five principles for AI governance in employment: human oversight of consequential decisions; transparency to affected workers; non-discrimination and bias auditing; data minimization and purpose limitation; and meaningful worker participation in governance. These principles, visible across the cases in this module, offer a practical checklist for evaluating any organization's AI labor practices.
The convergence of regulatory pressure (EU AI Act, EEOC, NLRB), collective bargaining precedent (WGA, SAG-AFTRA, Teamsters, UAW), and international frameworks (ILO, EU Platform Work Directive) suggests that employment AI is moving — unevenly and not without conflict — toward a world where worker consent, transparency, and joint governance are baseline expectations rather than exceptional demands. Organizations that build these structures proactively rather than waiting for compulsion are likely to face less disruption, litigation, and labor conflict than those that do not.
You are an HR director at a 3,000-person healthcare organization. Your company is planning to deploy an AI system that will: (1) screen job applications, (2) evaluate nurse performance based on patient data and shift logs, and (3) flag workers for additional review or disciplinary processes. Your works committee has asked for a governance framework before deployment. Work through it with the AI advisor.