Amazon's Atlas AI Union-Risk Scoring in UK Warehouses: What the Leaked Documents Show

Amazon uses an in-house AI forecasting tool called Atlas to predict how likely warehouse workers in the US, Canada and the UK are to unionise or take other collective action, according to internal documents made public as part of a court battle between Amazon and a Canadian union. The documents indicate Atlas assigns each warehouse a union-risk score between 100 and 900, and that two UK sites — a robotic fulfilment centre in Sutton Coldfield and the warehouse near Coventry — sit in the highest-risk tier.

The story was reported by openDemocracy, which reviewed the documents and interviewed Amazon workers, union organisers and lawyers about them.

What Atlas is and how the scoring works

Atlas is described in the documents as an in-house AI forecasting tool that predicts the likelihood of warehouse staff attempting to form a union, staging a sit-in, or carrying out other "disruptive labour events." The documents appear to be training resources for Amazon managers in Canada.

An appendix in the documents states that Atlas amasses "hundreds of data points" to award each warehouse a risk score of between 100 and 900 for union activity. Those data points include:

Posts on Reddit forums
Employees' responses to internal feedback surveys
Wider information about the workplace
Any previous industrial action at the site

The scoring direction is counterintuitive: the lower the score, the more likely Atlas believes workers — whom Amazon calls "associates" — are to unionise or take other actions that could affect Amazon's productivity or profitability. The document states that for Amazon leadership, higher scores are "better."

The five-tier risk scale

Warehouses appear to be categorised into five tiers based on their Atlas score. According to the documents:

Tier 5 means "no immediate perceived risk"
Tier 1 means "future activity is imminent"

The document suggests the scale can be used to "inform staffing decisions, resource planning, standard actions, and ownership to mitigate risks to Amazon labour structure."

Sixteen Amazon warehouses across the US, Canada and the UK are rated Tier 1, according to the documents. Two of those are in the UK: the £500m robotic fulfilment centre in Sutton Coldfield, north-east Birmingham, which Amazon refers to as EMA4, and the warehouse near Coventry, known as BHX4.

The Coventry warehouse and the 2024 union vote

Garfield Hylton, 62, works at the BHX4 warehouse near Coventry, receiving goods into the warehouse and dispatching them on to other warehouses. He is a member of the GMB union and has long known that Amazon monitors and logs every minute of his working day — every product he scans, every second between scans, and every period when his scanner is briefly inactive. That data is tracked, recorded, and used to set productivity targets and predictions.

What Hylton did not know, until GMB told him about Atlas, was that Amazon seemingly also monitors potential "dissent" at his warehouse using AI to surveil social media and the company's internal staff bulletin board.

"It's not right in a democracy… the employer should not be interfering. Workers should be allowed to decide if they wish to choose to unionise. If you're then monitoring, you're undermining a process deliberately and willfully." — Garfield Hylton, Amazon warehouse worker and GMB member

GMB has long tried to unionise Amazon workers in the UK. Employees at the Coventry warehouse became the first in the country to take strike action against Amazon in January 2023. Around 2,600 workers took part in an official ballot for union recognition at the site six months later, which GMB lost by just 28 votes.

Rachel Fagan, an organiser for GMB, said the documents confirm what the union suspected.

"It's quite clear from these documents that these sites are being monitored for union activity. Everyone knows that Amazon is heavy on surveillance, but this becomes more personal. Amazon is using this technology to essentially outwit and manoeuvre workers that want to organise. Nothing in this fight against Amazon is an even playing field; they have all the wealth, all the technology, everything to their disposal and they use it." — Rachel Fagan, GMB organiser

The "dirty tricks" allegations around the Coventry vote

GMB and Foxglove, a tech justice non-profit, have accused Amazon of using "dirty tricks" to swing the Coventry recognition vote in its favour. The allegations include mass hiring of non-unionised workers and allegedly plastering QR codes over screens, cafeteria tables and toilet doors in the warehouse in the run-up to the vote. Employees could scan those codes to cancel their GMB membership.

Foxglove and the law firm Leigh Day are representing two workers in a legal challenge against Amazon's use of these "click-to-quit" QR codes. They argue the tool was an attempt to "bully" workers into ending their union membership.

Rosa Curling, co-executive director of Foxglove, described the level of surveillance detailed in the documents as "highly intrusive," showing that Amazon has a "very intentional policy to try and bust organising, bust unionising efforts, and they are using all sorts of tools to do that."

"We are actively reviewing these documents to see what additional arguments, if any, we should be adding to our existing claim to unlawful inducement, or indeed on this ongoing breach of workers' rights to organise, which is a well-respected right in UK-made laws." — Rosa Curling, co-executive director of Foxglove

Michael Newman, a partner at London-based law firm Leigh Day, told openDemocracy that the documents make it clear Amazon views unions as bad, and that this tells "a very different story" from Amazon's previous assertions that it does not seek to prevent workers from joining a union.

"Now we know Atlas is used by Amazon, we'll be asking more questions about what data points are used, how these scores were reached." — Michael Newman, partner at Leigh Day

Where the documents came from

The documents revealing Amazon's Atlas AI tool were disclosed as part of a legal challenge by the Canadian union Unifor and shared with the US non-profit publication Truthout. The British Columbia Labour Board ultimately ruled that Amazon had violated multiple local labour laws, following an earlier ruling that the company had unlawfully imposed a wage freeze on unionised workers at a warehouse outside Vancouver.

Amazon's response

Amazon refuted the characterisation of Atlas as "individual employee monitoring or surveillance," stating: "This is site/building level activity only."

The company said it does "not track individual employees' opinions, affiliations or union membership," or any "personal data," and said Atlas is not used to monitor dissent but to "help us understand where associates need more support and to ensure we're addressing issues that they raise."

Amazon also told openDemocracy that the Atlas framework is not currently used in the UK. The company had not responded to openDemocracy's question about whether it uses a similar system under a different name at the time of publishing.

In a fuller statement, an Amazon spokesperson said:

"The GMB's characterisation is wrong. This isn't surveillance. Like any responsible employer, we have numerous ways for our team to provide us with feedback, and we use that feedback to identify where we can improve the day-to-day experience of our employees. Our employees always have the choice of whether or not to join a union. What matters most to us is the direct relationship between our teams and their managers — it means every person's voice is heard, not just a few, and that when someone spots a way to improve their job, we can act on it quickly. We've made real progress in recent years on the things that matter, like pay and safety, and we know there's always more to do. That's our focus: getting better every day."

The UK government's position

A government spokesperson did not endorse or condemn Atlas specifically, but set out general expectations for workplace monitoring:

"Employers and workers have a shared interest in workplace monitoring technologies being used transparently and in ways that are fair, accountable and reliable, so that they can support productivity and good management while maintaining trust in the workplace. We expect all employers to respect workers' rights to take part in lawful trade union activity."

Amazon's prior surveillance record

Amazon has previously been found guilty of "excessive" surveillance of its workers. In 2022, the French data watchdog CNIL fined the company €32m (£27m) for breaching General Data Protection Regulation (GDPR) rules by tracking worker activity so precisely that an alert was triggered if workers were inactive for between one and 10 minutes.

What the discussion around the story highlighted

Commenters on Hacker News focused on three themes: the chilling effect of scoring workers' opinions, the gap between how such systems are marketed internally and what they actually measure, and the broader asymmetry between corporate and state surveillance.

One commenter argued that no form of mass surveillance is defensible on balance:

"There is no form of surveillance that does more good than harm in the long run. The endless debates about the value of 'catching bad guys' or 'keeping people on their best behavior' should be rightly seen as opposed to the fundamental notion that human beings are capable of cooperation and altruism at scale. No amount of mass surveillance is justifiable, and there is no version of the future in which we allow it but do not create dystopia."

Another commenter pointed to the internal survey question shown in the reporting — "I am optimistic about the future of AI at Amazon" — and observed that any answer other than "Strongly Agree" would very likely flag the respondent as a risk.

A commenter who said they had worked in workforce analytics at a large logistics company described how these systems are framed internally:

"Did workforce analytics at a big logistics outfit and this stuff is usually sold internally as 'attrition risk' or 'engagement heatmaps', never as union anything. Same features though, shift churn, grievance counts, break room badge patterns."

One commenter questioned the technical premise, asking whether there is enough training data for the monitoring results to be close to accurate. Another framed the core objection as one of protected freedom:

"I will never understand how corporations are allowed to interfere in any way with unions. It should be a protected freedom to associate and expression. Why is it allowed to be gameified where corporations can detect it early and stop it? Hands-off people's lives — unions aren't illegal, so no interference."

A separate thread compared corporate surveillance with government surveillance, with one commenter arguing that government surveillance is worse but that European countries focus on restricting only the corporate kind. That commenter also cited Amazon UK's requirement of ID and a selfie for "bladed items" and a ranking that places the UK among the least free countries for digital freedom. Other commenters responded to the story with shorter reactions, including one who said they had closed their Amazon UK account and another who described the situation ironically as a "healthy work environment."

Key facts at a glance

Tool: Atlas, an in-house Amazon AI forecasting tool
Purpose as described in the documents: predict likelihood of union formation, sit-ins and other "disruptive labour events"
Score range: 100 to 900 per warehouse; lower scores indicate higher perceived union risk
Inputs: hundreds of data points, including Reddit posts, internal feedback survey responses, workplace information and prior industrial action
Tiers: five, from Tier 5 ("no immediate perceived risk") to Tier 1 ("future activity is imminent")
Tier 1 sites: 16 across the US, Canada and the UK, including two in the UK — EMA4 in Sutton Coldfield and BHX4 near Coventry
Amazon's position: site-level activity only; not individual surveillance; Atlas not currently used in the UK
Prior penalty: €32m (£27m) CNIL fine in 2022 for GDPR breaches related to worker activity tracking