← Back to feed

Internal training documents unsealed during a Canadian labour tribunal have revealed that Amazon operates an AI system called Atlas that assigns warehouses a union-risk score between 100 and 900. Lower scores indicate a higher probability of workers attempting to organise. Two UK sites are currently rated Tier 1 — the highest-risk category, defined in the documents as “future activity is imminent.”

What Atlas Tracks

The documents, filed as part of a court battle between Amazon and a Canadian union, describe Atlas as an in-house AI forecasting tool that aggregates “hundreds of data points” per site. Inputs include:

• Posts on public Reddit forums
• Employee responses to internal feedback surveys
• Amazon’s internal staff bulletin board activity
• Workplace demographic and operational data
• History of prior industrial action or organising attempts

Atlas produces a composite score that feeds into a five-tier classification system. Tier 5 means “no immediate perceived risk.” Tier 1 means action is considered imminent. The document states the score is intended to “inform staffing decisions, resource planning, standard actions, and ownership to mitigate risks to Amazon labour structure.”

Of 16 warehouses globally rated Tier 1, two are in the UK:

• BHX4 — a fulfilment centre near Coventry, site of the first formal strike by Amazon workers in the UK (January 2023) and a narrow union recognition ballot loss by 28 votes in mid-2024
• EMA4 — a £500m robotic-fulfilment centre in Sutton Coldfield, north-east Birmingham

Amazon’s Response

Amazon disputed the characterisation of Atlas as individual employee surveillance. A company spokesperson said the system operates at the “site/building level only” and denied that personal data, union membership, or individual opinions are tracked. The company framed Atlas as a tool “to help us understand where associates need more support and to ensure we’re addressing issues that they raise.”

The documents, however, describe uses inconsistent with that framing: Atlas inputs explicitly include social media monitoring for “dissent,” and the scoring scale is described as a mechanism for management to “mitigate risks,” with higher scores described as “better” for company leadership.

Why It Matters

The revelations surface a specific, documented use of AI in labour management that goes beyond productivity tracking. Models trained on workers’ online speech and survey responses — to predict and preempt collective action — raise questions about the lawfulness of such systems under UK data protection rules, and the asymmetry of information between a company with access to predictive AI and workers trying to organise.

GMB union organiser Rachel Fagan called it “quite clear from these documents that these sites are being monitored for union activity” and said the technology creates an uneven playing field: “They have all the wealth, all the technology, everything to their disposal and they use it.”

Michael Newman, a partner at Leigh Day law firm, said the Atlas documents tell “a very different story” from Amazon’s prior public statements on worker organising.

The UK’s Information Commissioner’s Office has not yet commented on whether Atlas-style social media scraping for union-risk scoring is compliant with the UK GDPR.