Amazon has been accused of using an artificial intelligence-powered system to assess the likelihood of union organising at its warehouses in the United Kingdom, United States and Canada, according to documents disclosed during legal proceedings.
The system, known as Atlas, reportedly assigns warehouses risk scores based on the perceived likelihood of labour organising and workplace disruption, according to an investigation by openDemocracy.
The allegations emerged from a document released as part of court proceedings involving Amazon and Canadian trade union Unifor. The document, which appears to serve as a training manual for company managers, describes Atlas as a tool for evaluating labour-related risks across the company's operations.
The findings have prompted criticism from Britain's GMB Union, which accused Amazon of deploying “Orwellian” technology to monitor its workforce.
According to the document, Atlas assigns warehouses a score ranging from 100 to 900, with lower scores indicating a greater perceived likelihood of union activity.
Higher scores are described as preferable for company management.
Michael Newman, a partner at London-based law firm Leigh Day, told openDemocracy that the scoring system “makes it clear that they think the union is bad”.
The document describes Atlas as a “labour risk framework to clarify how Amazon evaluates emerging workforce issues, labour union targets and internal or external labour threats.”
It also establishes a five-tier classification system for assessing workplace risks.
Under this framework, tier one represents facilities experiencing disruption or facing imminent worker action, while tier five indicates locations where there is no immediate perceived risk.
The classifications can reportedly inform staffing decisions, resource allocation and measures intended to mitigate risks to productivity and profitability.
Two Amazon warehouses in the United Kingdom were placed in the highest-risk category, alongside 14 facilities in the United States and Canada.
Among the British facilities identified was BHX4, near Coventry, where workers have been involved in an active union organising campaign.
Another facility, EMA4 in Sutton Coldfield, was also included in the highest-risk category following previous strike action.
The revelations have raised concerns among labour representatives about the potential use of workplace data and automated assessment systems to identify and respond to collective organising efforts.
However, Amazon rejected allegations that Atlas was designed to undermine trade unions or monitor individual employees' political or labour affiliations.
An Amazon spokesperson told openDemocracy that the company does not “track individual employees’ opinions, affiliations or union membership”.
The spokesperson also denied that Atlas constituted surveillance of individual workers or was used to counter union activity.
According to Amazon, the system is intended to “help us understand where associates need more support and to ensure we’re addressing issues that they raise”.
The company maintained that its purpose is to identify workplace concerns and improve responses to employee feedback.
The disclosed material indicates that Amazon evaluates labour-related risks at the warehouse level, although it does not establish that the company uses Atlas to predict individual employees' union membership or intentions.
The controversy highlights growing scrutiny of how major employers use artificial intelligence and workforce analytics in managing labour relations.