By Rameen Sajjad, Fall 2025 Marcellus Policy Fellow

American technology companies face growing criticism for extracting data and digital labor particularly through data annotation from developing nations in Africa and Asia. The extraction is done without equitable compensation or adequate regulatory oversight. This unregulated resource extraction model is characterized by scholars as digital colonialism, which breeds resentment and undermines U.S. credibility in regions vital to the stability of the U.S. AI supply chain, potentially creating openings for rivals like China. The pursuit of AI dominance has diverted policymakers’ attention from enforcing ethical standards on technology giants. Exploitation manifests in financial abuses and exposure to graphic content.

The boom of generative AI has intensified scrutiny on this hidden infrastructure of invisible workers, revealing that AI’s intelligence relies on massive, low-cost human effort primarily sourced from the Global South. U.S. tech giants often use opaque subcontracting chains through third-party vendors to distance themselves from terrible working conditions, obstructing meaningful accountability. This short-term exploitation to win the AI race jeopardizes the United States’ long-term strategic goals. Corporate exploitation triggers blowback in host nations and it can manifest in multiple forms including high turnover, talent flight to competitors offering better compensation, stagnation in innovation, and even geopolitical risks such as looming data embargoes. Current US policy contains significant gaps. The Trump Administration rescinded its predecessor’s mandates for ethical standards (EO 14110) in 2025 in favor of deregulation focused on innovation acceleration and speed, while existing export controls and Department of Labor initiatives overlook labor ethics in global annotation pipelines.

To safeguard critical AI supply chains and prevent key data-rich nations from aligning with Chinese ecosystems, the United States must mandate transparency and accountability for the global operations of domestic tech firms. This strategy is essential to preempting anti-U.S. data restrictions and maintaining a competitive digital edge. The proposed policy framework aims to transform exploitative practices into reliable and reciprocal partnerships that deliver three strategic benefits to Washington: (1) a stable, high-quality labor pipeline; (2) reduced bias and superior performance of U.S. AI systems through motivated, better-supported workers; and (3) strengthened alliances in the Global South that block Chinese inroads. This framework includes three core mandates: Mandatory Ethical Audits, requiring annual third-party inspections of AI supply chains worldwide, built upon a hybrid standard incorporating International Labor Organization (ILO) conventions and established risk management frameworks. Second, Fair Labor Certifications should tie U.S. export licenses to verifiable global labor standards, including living wages, mental-health safeguards, and a total ban on non-disclosure agreements (NDAs) that silence victims. Third, the government should mandate Metrics for Equitable Partnership Mechanisms (MEPM), requiring U.S. firms to disclose quantifiable indicators of local, in-country AI investment (such as R&D hubs and training centers) to ensure outsourcing actively builds sovereign AI capacity in partner nations like Kenya, India, and the Philippines. Moving beyond extractive data practices ensures a stable, ethical labor supply chain that is critical for training high-performing AI systems which are essential for economic prosperity and national security. Strengthening the AI commercial base creates a robust dual-use ecosystem, allowing private-sector innovation to serve as a reliable engine for national security.

Continued inaction risks high turnover and talent flight to Chinese firms, degradation in training-data quality, innovation stagnation, and escalating geopolitical blowback — including restrictive data-sovereignty laws and outright data embargoes already emerging in Kenya and India that could choke off the diverse, high-volume datasets American models critically depend on for its security and economic prosperity needs.