Summary
Press release from Alphabet Workers Union-CWA reporting on a survey of 160 US-based data workers, documenting that exploitative conditions seen in the Global South also exist within the domestic AI supply chain. Produced in partnership with TechEquity and the Communications Workers of America.
Key Points
- 86% of surveyed workers worry about meeting financial responsibilities; 25% rely on public assistance (food assistance, Medicaid)
- Median hourly wage: $15; median paid workweek: 29 hours; median annual earnings: ~$22,620
- 66% spend at least 3 hours weekly waiting for tasks (unpaid); 26% spend 8+ hours waiting
- Only 30% are paid for time when no tasks are available
- 87% regularly assigned tasks for which they are not adequately trained
- Only 23% have employer-provided health insurance
- 52% believe they are training AI to replace other workers’ jobs; 36% believe they are training AI to replace their own jobs
- 74% concerned about AI’s contribution to disinformation spread; 54% concerned about surveillance
- Over half felt Average Estimated Times (AETs) for tasks were too short to complete accurately — meaning rushed work increases AI system risk
- AWU conclusion: “While Big Tech rakes in the profits from the AI boom, workers are struggling to make ends meet”
Newsletter Angles
- The domestic angle: this is not just an offshore labor problem. US workers building AI systems for Google are on public assistance. The political economy of AI includes subsidizing it with social safety net dollars
- The self-replacement trap: 52% of data workers believe they’re training AI to replace others; 36% believe they’re training AI to replace themselves. Workers can see the endgame clearly — they have no alternatives
- Training data quality is a function of working conditions: workers rushed by inadequate AETs produce lower quality training data. The exploitation doesn’t just harm workers — it degrades AI system quality
Entities Mentioned
- Amazon — Mechanical Turk mentioned as platform model
- Meta / OpenAI — implicit employers via contractor chains (Google/Alphabet as the direct context for AWU)
Concepts Mentioned
- Mechanical Turk Pattern — US data workers face the same structural conditions as offshore workers; the pattern is universal, not geographical
- Leverage Erasure Through Automation — workers explicitly recognize they are building their own replacements
Quotes
“The solution to this organized greed is organized labor.” — Parul Koul, President of AWU-CWA Local 9009
“What has changed due to the rise of AI is the scale at which technology fragments and obscures work and its conditions. Layers upon layers of outsourcing workers enable tech companies to try to outsource their liability.” — Catherine Bracy, TechEquity
Notes
This is a press release from an advocacy organization (AWU), so the framing is explicitly pro-labor. The data (survey of 160 workers, 15 in-depth interviews) is presented without full methodology in this press release; the full report is linked separately. The conclusions are directionally consistent with independent academic research on ghost work conditions.