The dark side of AI: exploitation and bias behind the algorithms

Artificial intelligence is reshaping daily life, but its development often relies on hidden human costs. Generative AI and large language models now power tools like chatbots, yet their training depends on underpaid workers in harsh conditions. These systems also risk reinforcing biases, as seen in policing and social services. In November 2022, OpenAI launched ChatGPT, a chatbot driven by a large language model. Its release marked a turning point for AI’s role in everyday tasks. Yet behind such tools lies a less visible workforce.

In 2021, concerns over discrimination led to a pause in some machine-learning programmes. Denmark and Australia faced criticism for using biased algorithms to judge benefit eligibility. US law enforcement has also employed facial recognition and predictive policing, with disproportionate impacts on communities of colour.

OpenAI partnered with Sama, a US-based firm, to recruit Kenyan workers for data labelling. They earned between $1.32 and $2 per hour. Many reported severe trauma from moderating extreme content. Millions globally perform similar tasks under poor conditions to train AI systems.

Despite this, tech leaders often portray AI as self-sufficient. The reality is that human labour remains essential, though frequently exploited. The rise of AI brings both innovation and ethical concerns. Exploitative labour practices and algorithmic biases demand scrutiny. Without careful oversight, these technologies risk perpetuating colonialism, exploitation, and systemic inequality.