Giant artificial intelligence (AI) firms such as OpenAI and Meta face new minimum pay and mental healthcare rules in Kenya as the government seeks better working conditions for local staff, including content moderators and data annotators.
A proposed policy by the ICT Ministry says the government is developing protection guidelines for moderators, who review and remove harmful material from online platforms, and annotators, who label images, text and audio to train AI models such as ChatGPT to recognise and respond to human prompts.
The guidelines will require AI companies and outsourcing firms to comply with locally set duty-of-care standards, including safeguards against harmful content, access to mental health support and transparent contracting practices.
The government will publish occupational protection guidelines covering minimum standards for written contracts, psychosocial support, grievance mechanisms and working conditions.
For years, Kenyan workers employed by outsourcing firms serving global technology companies such as ChatGPT owner OpenAI and Facebook parent firm Meta have complained of psychological trauma and unfair payment.
“Support the development and integration of fair and transparent pay standards for AI and other Emerging Technologies value chain workforce,” reads the draft policy.
“Promote the development, enforcement, and compliance with duty-of-care standards for AI and other Emerging Technologies value chain workers, including safeguards against harmful content exposure, access to mental health support, transparent contracting practices, proportionate workplace surveillance measures, and accessible grievance and redress mechanisms.”
The proposed policy further says a fair-pay-reference framework will set transparent pay benchmarks for data annotation, content moderation and AI quality evaluation roles, calibrated against international rates for equivalent work.
Companies employing Kenyan AI workers would also be required to disclose their pay structures against those benchmarks through a compliance reporting mechanism.
In the last five years, Kenya has emerged as a global hub for AI data annotation and content moderation because of its large English-speaking workforce.
Moderation and annotation are critical to the development of generative AI systems such as ChatGPT, Gemini and Microsoft’s Copilot, as they rely on human reviewers to help train algorithms to recognise prompts that could generate harmful content.
The algorithms behind these bots rely on vast amounts of human-labelled data to identify harmful content and improve their responses.
Technology companies increasingly outsource the work to specialist contractors in countries such as Kenya to reduce labour costs while creating legal distance from the employment relationship.
By outsourcing these services, tech giants significantly slash expenses by paying significantly lower wages compared to hiring domestic workforces in the US or Europe.
Markets like Kenya, India, and the Philippines have high youth unemployment, creating a large, eager pool of workers who will accept low pay.
Using external vendors also allows tech giants to distance themselves from direct responsibility for worker welfare and compensation.
Kenyan moderators and annotators working on projects for Meta and OpenAI through outsourcing company Sama, for instance, have previously raised concerns over severe psychological trauma, low pay and abrupt layoffs.
The workers say they are exposed to graphic violence, self-harm, murder, child abuse, rape, necrophilia, bestiality and incest, as well as deeply invasive, non-consensual personal video footage captured by Meta AI smart glasses, while alleging they received little or inadequate psychological support from the US-headquartered firm.
Sama has denied the allegations.
Some content moderators were paid between $1.46 (Sh189) and $3.74 (Sh484) an hour. In the US, moderators are paid an average of $21 (Sh2,719) to $27 (Sh3,496) per hour.
“Data annotation and content moderation workers face unique occupational risks, including exposure to harmful content, insecure working conditions, and limited labour protections,” Kenya’s draft policy says.
“Existing frameworks provide insufficient safeguards for mental wellbeing, transparency, and employer accountability.”
The draft policy says the guidelines apply to local and international AI firms operating in Kenya.
It would be one of Africa’s most comprehensive labour protection frameworks specifically targeting AI value-chain workers.
