The Dark Side of AI: Exploitation and Bias Behind Rapid Growth

Generative AI and large language models now play a growing role in daily life. Yet behind these technologies lies a hidden workforce facing harsh conditions. Millions of underpaid labourers perform repetitive tasks to label data, often under exploitative terms.

The rapid expansion of AI has also raised concerns about bias and discrimination. Some cities and countries have already restricted its use due to unfair outcomes, particularly against marginalised groups. OpenAI launched ChatGPT, a chatbot powered by a large language model, in November 2022. The system relies on vast amounts of labelled data, much of it processed by low-wage workers. To meet demand, OpenAI partnered with Sama, a US-based firm that hired Kenyan data labellers for as little as $1.32 an hour.

Workers at Sama have reported severe trauma from exposure to extreme content while labelling data. The company has faced criticism for its labour practices, which critics argue exploit vulnerable employees. No exact global figure exists for the number of people in such 'ghost work' roles, but the scale is believed to be vast.

Beyond labour concerns, AI systems have shown repeated biases. US law enforcement has used facial recognition and predictive policing tools, disproportionately targeting communities of colour. Cities like San Francisco and Boston have banned facial recognition software after cases of misidentification. Similar issues arose in Denmark and Australia, where AI-driven welfare assessments produced discriminatory results. Some governments have paused machine-learning programmes over these risks.

Experts warn that without scrutiny, AI development could reinforce colonialism, exploitation, and systemic ignorance. The push for progress, they argue, must not come at the expense of fairness or human cost. The rise of AI technologies has exposed deep ethical challenges. From underpaid data labellers to biased algorithms, the risks of unchecked expansion are clear. Without stronger safeguards, these systems may continue to harm the very people they claim to serve. Some regions have already taken action, but broader reforms remain uncertain.