Meta’s AI training relies on employee surveillance: systemic exploitation of labor to fuel extractive data capitalism

Indigenous epistemologies reject the commodification of human activity as 'data,' framing it as a violation of relational trust and collective sovereignty. Many traditions treat observation and labor as sacred, not extractable resources—contrasting sharply with Meta’s reduction of human behavior to training inputs. The concept of 'data sovereignty' in Māori and Aboriginal legal frameworks asserts that knowledge cannot be owned or surveilled without consent, a principle systematically ignored in Silicon Valley. This dimension scores low because the narrative reflects no engagement with these traditions.

The use of employee tracking mirrors historical management techniques like Taylorism (scientific management) and Fordism, which treated workers as cogs in a machine to maximize output. These systems were later critiqued for dehumanizing labor and enabling exploitation under industrial capitalism. The digital era extends this logic, replacing stopwatches with keystroke logs and mouse movements, but the power imbalance remains. The shift from factory floors to remote workspaces does not eliminate coercion—it merely obscures it behind algorithmic control.

Cross-culturally, the tension between data as property (Western) and data as commons (many Indigenous and Global South societies) is stark. In India, the Right to Privacy judgment (2017) recognized surveillance as a threat to dignity, while in the EU, GDPR attempts to limit corporate data extraction. Meanwhile, China’s surveillance state demonstrates how tracking can be weaponized for social control. Meta’s approach aligns with the former, treating data as a private asset to be mined, rather than a shared resource to be stewarded.

From a scientific standpoint, training AI on employee behavior data risks introducing systemic biases, as the data reflects the power dynamics of the workplace rather than neutral human activity. Studies show that surveillance increases stress and reduces creativity, undermining the very 'high-quality' data Meta claims to seek. The methodology also ignores the ecological footprint of training AI on such datasets, which require massive computational resources. Peer-reviewed research on algorithmic management consistently warns of these pitfalls, yet industry practices remain unchecked.

Artists and spiritual traditions often critique the reduction of human experience to quantifiable data, framing it as a form of violence against the ineffable. Writers like Ursula K. Le Guin and Octavia Butler have explored the dehumanizing effects of surveillance societies, while Indigenous ceremonies treat observation as a reciprocal act, not an extractive one. The act of tracking keystrokes and mouse movements strips away context, reducing life to a series of inputs—a spiritual desecration of human agency. This dimension is largely absent in the original narrative, which treats data as inert.

If unchecked, this model risks normalizing total workplace surveillance, where every action is monetized and every worker is a data point. Future scenarios include AI systems trained on increasingly intimate personal data (e.g., biometrics, emotional states), exacerbating power asymmetries between employers and employees. Regulatory frameworks like the EU AI Act may curb some excesses, but without structural changes, surveillance capitalism will deepen inequality. The long-term implication is a world where human labor is not just exploited but algorithmically optimized into obsolescence.

Marginalized workers—gig workers, outsourced staff, and those in Global South subsidiaries—are most vulnerable to such surveillance, as they lack labor protections or bargaining power. Women and people of color, who are overrepresented in precarious tech roles, face heightened risks of discrimination via algorithmic management. The narrative erases these voices, framing the issue as a neutral 'technical challenge' rather than a systemic injustice. Without their inclusion in policy discussions, solutions will perpetuate existing hierarchies.