The Labour of AI

Key Takeaways

• AI models rely on gig‑scale “ghost work” performed by low‑paid workers
• OpenAI’s partnership with Sama in Nairobi exemplifies outsourced data labeling
• Tools often obscure labor details, citing “ethical supply chains” or refusing answers
• Wittgenstein’s 1947 insight frames AI as “humans who calculate”
• Transparency gaps risk reputational and regulatory challenges for AI firms

Pulse Analysis

The rise of generative AI has been powered by a sprawling, low‑visibility labor market often termed "ghost work." Hundreds of thousands of micro‑tasks—image tagging, content moderation, transcription—are outsourced to gig platforms that pay per click, sometimes below living‑wage thresholds. This model enables rapid scaling of training data while keeping costs low, but it also creates a supply chain that is difficult to audit and prone to exploitation. Scholars such as Ruggiu and Özdemir have documented how these invisible workers form the backbone of AI performance, turning human cognition into a commodity.

A concrete illustration of this dynamic is the partnership between OpenAI and the Nairobi‑based social enterprise Sama. Through Sama, OpenAI contracts local workers to label and curate massive datasets, a process essential for refining large language models. While the collaboration promises economic opportunities in Kenya, reports indicate that wages remain modest and job security limited, echoing broader concerns about digital slavery. The case highlights how AI firms can outsource ethically fraught work to regions with weaker labor protections, raising questions about corporate responsibility and the true cost of AI innovation.

The opacity of AI labor pipelines poses strategic risks for businesses. Investors are increasingly scrutinizing ESG metrics, and regulators are exploring legislation that could mandate supply‑chain transparency for high‑impact technologies. Companies that proactively disclose their data‑labeling practices, invest in fair‑pay standards, and embed worker welfare into model development will likely gain a competitive edge. As the market matures, clear guidelines and third‑party audits could become the norm, turning today’s hidden labor into a measurable component of AI’s total cost of ownership.