A look at a new labor force using microtask platforms like Remotasks to create and edit data sets for self-driving cars, located primarily in the Global South
Context & Ripple Effects
This piece is the origin point for a labor beat that kept developing: it documents how platforms like Remotasks mobilize gig workers across the Global South to create and edit the training datasets self-driving cars depend on. Two years later, interviews with 36 Remotasks workers in the Philippines found the model's weak spot — delayed, reduced, or canceled payments for annotation work done for Scale AI's US customers.
The coverage since has widened the frame from self-driving data to a general geography of cheap AI labor: [[a:839906|US tech companies hiring in South and Central America for time-zone-compatible, lower-cost remote work]], Chinese firms routing Kenyan labelers through opaque middlemen on WhatsApp, and Indian workers paid to wear first-person video devices that turn physical work into robot training data.
First-order effects
- Remotasks and Scale AI get a flexible annotation workforce whose output directly feeds self-driving programs, while workers in the Global South gain income tied to opaque task pricing they do not set.
- US self-driving customers effectively outsource the most labor-intensive step of dataset creation to a workforce with little bargaining power or visibility.
Second-order effects
- Payment disputes of the kind later reported in the Philippines push workers and observers to scrutinize platform accountability, pressuring buyers like Scale AI to police how their labeling vendors pay.
- The arbitrage normalizes hiring AI-adjacent labor wherever it is cheapest, pulling adjacent markets — freelance design platforms and LatAm remote hiring — into the same cost competition.
Third-order effects
- If the pattern holds, AI's supply chain consolidates into an invisible labor tier spanning annotation, middleman networks, and even wearable data capture, with accountability problems recurring faster than labor protections catch up.
- The same Global South workforce that labels today's models is also the most exposed to being replaced by the systems it trains, as the freelancer coverage suggests.
The trend: AI development is building a distributed, largely unseen Global South labor force whose conditions and pay disputes are becoming the industry's accountability frontier.