Artificial intelligence companies operating in Kenya could soon be compelled to improve how they treat local workers, with the government proposing new rules that seek fair pay, mental health support and stronger workplace protections for thousands of people powering the global AI industry.
The Ministry of Information, Communications and the Digital Economy has unveiled a draft policy that seeks to protect workers involved in AI and digital platforms, particularly content moderators and data annotators.
These workers are responsible for reviewing harmful online material and preparing data used to train AI tools such as ChatGPT to better understand and respond to human prompts.
Under the proposal, AI companies and outsourcing firms will be expected to meet locally developed duty-of-care standards covering worker safety, fair treatment and wellbeing. The policy also seeks to ensure employees exposed to disturbing online content receive mental health support while promoting clear employment terms and better workplace practices.
The government further plans to introduce occupational protection guidelines that will set minimum standards for written contracts, psychosocial care, complaint handling systems and working conditions for people employed across the AI value chain.
The move comes after years of complaints from Kenyan workers hired by outsourcing firms serving global technology companies, including ChatGPT owner OpenAI and Facebook parent company Meta. Many have raised concerns over the emotional impact of reviewing disturbing content as well as poor pay.
The draft policy states: “Support the development and integration of fair and transparent pay standards for the workforce in the value chain of AI and other emerging technologies.”
It further states: "Promote the development of, enforcement of, and compliance with duty-of-care standards for AI and other emerging technologies value chain workers. These standards include safeguards against harmful content exposure, access to mental health support, transparent contracting practices, proportionate workplace surveillance measures, and accessible grievance and redress mechanisms."
The policy also proposes the creation of a fair-pay reference framework that will establish open pay benchmarks for workers involved in data annotation, content moderation and AI quality evaluation. The benchmarks will be based on what workers carrying out similar duties earn in other countries.
Companies hiring AI workers in Kenya will also be expected to disclose how their salaries compare with those benchmarks through a compliance reporting system.
Kenya has grown into one of the world's preferred destinations for AI data annotation and content moderation over the past five years, largely because of its large English-speaking workforce.
The work carried out by content moderators and data annotators is central to the development of generative AI systems such as ChatGPT, Gemini and Microsoft's Copilot. They review and label text, images and other content to help AI models identify harmful material and improve the quality of their responses.
These AI systems rely on massive amounts of human-reviewed data to learn how to respond safely and accurately to users.
Many global technology companies contract specialist firms in countries such as Kenya to perform this work, allowing them to lower labour costs while keeping a distance from direct employment responsibilities.
The arrangement has enabled technology firms to spend less on wages than they would if they employed workers in the United States or Europe. Kenya, India and the Philippines have become attractive outsourcing destinations because they have large numbers of young people looking for employment.
The outsourcing model has also allowed technology companies to reduce their direct responsibility for employee welfare and compensation.
Some content moderators have earned between $1.46 (Sh189) and $3.74 (Sh484) an hour. In comparison, moderators in the United States receive an average of between $21 (Sh2,719) and $27 (Sh3,496) an hour.