• Amazon has reportedly been using an AI-powered algorithm to monitor union activity at sites around the world
• Two sites in the UK were recorded at the highest risk level for union activity, raising questions over the legality of the AI algorithm

Amazon has been embroiled in a legal dispute over unionization at a warehouse just outside of Vancouver, and while that may not sound particularly significant, it has forced the company to turn over documentation to the British Columbia Labour Board and Unifor.

These documents have revealed that Amazon has deployed an AI-powered union activity detection algorithm to assess whether its workers at warehouses around the world are at risk of unionizing.

This AI algorithm - named Atlas within the documentation - uses hundreds of data points to place warehouses on a unionization risk scale. These data points include worker communications channels, health and safety incidents, minimum wage rates, and even the proximity of a non-unionized warehouse to one exposed to union activity.

UK union activity under surveillance

Outside of monitoring Amazon’s internal ‘Voice of the Associate’ bulletin board, it has been revealed that Atlas also monitored external forums used by workers such as Reddit to “to understand the tenor of the discussion at the site level,” the documents state.

The algorithm would monitor common phrases and keywords alongside upvotes and fellow worker responses to identify emerging risk of unionization.

Warehouses are then assessed on a risk scoring, with Tier 5 sites assessed as having little to no risk of union activity, and Tier 1 sites assessed as having imminent risk of union activity.

The activities of workers are also categorized in a risk system, with activities such as signing union cards and workers approaching management with demands deemed to signal increased risk, alongside organized actions such as protests, sickouts, and work stoppages.

The documentation also includes an over-100 page management training guide on union activity, how it can progress, and how it can affect the workplace - alongside how management should engage with employees discussing union activity.

While the documentation focuses mainly on warehouses in North America, two UK warehouses based in Coventry and Minworth whose employees have been working with GMB Union were designated as ‘Tier 1’ sites - meaning they are at imminent risk of unionizing.

“This is a shocking revelation,” said Rachel Fagan, GMB Organiser. “Amazon, one of the globe’s wealthiest corporations, could be deploying Orwellian AI-fuelled tools to monitor workers right here in the UK.”

“We’ve long warned that Amazon top brass is uninterested in the norms and expectations of British industrial relations. We’ve written to the Government demanding urgent answers. Workers deserve to be able to work free of the fear of excessive surveillance. If Amazon bosses are using Atlas or similar AI-fuelled systems in the UK, the British public deserve to know,” Fagan added.

GMB Union has also called on the Government to urgently seek answers from Amazon about the use of Atlas in British workplaces.

reached out to Amazon for comment, but did not immediately receive a response.

Is Amazon’s surveillance of UK warehouses illegal?

Under the Trade Union and Labour Relations (Consolidation) Act 1992, workers in the UK have the right to join a union and cannot be discriminated against for their intention to join a union.

These regulations appear not to have been violated. But where questions remain is over how Amazon processed the data it collected on workers from the Coventry and Minworth sites.

“I think from my perspective the allegations are concerning if true,” Jonathan Armstrong, Partner at Punter Southall Law, told TechRadar Pro. “In the UK the processing of data relating to trade union membership counts as special category data under GDPR Article 9. This means that extra care must be taken when processing that data.”

“Processing personal data using AI is also subject to the data protection principles in GDPR so processing must be lawful, fair and transparent. In most cases this will mean that employees should be told when their personal data is going to be processed by AI with a clear and transparent notice setting out what the plans are. In most cases there should also be a right to opt-out too,” Armstrong continued.

“In addition when bringing in a new AI system an employer should do a Data Protection Impact Assessment to work out whether it can put in place sufficient safeguards – if it can’t it may need to consult with a data protection regulator such as the Information Commission. And we have had cases which have highlighted to issues with processing excessive employee data.”

“For example in Germany H&M were fined €35.3m in 2020 for excessive processing of employee data to assist management in making employment related decisions. It seems that there are warehouses in the EU also involved here. If that’s the case Amazon will also need to consider the impact of the EU AI Act in addition to GDPR.”

“These are issues which employers need to take seriously – just because AI can do things that doesn’t mean it should. AI is not the wild west and whenever a system like this is contemplated the board need to make sure that the right legal analysis has been done and that the necessary legal protections are in place,” Armstrong concluded.