Kenya wants to know what the people who label AI training data are paid — and how that compares with what the same work earns elsewhere.
A draft national AI policy published by the ICT ministry would create a Fair Pay Reference Framework setting benchmark rates for three jobs at the bottom of the AI supply chain: data annotation, content moderation, and evaluating AI output for quality. The benchmarks would be “calibrated to international rates for equivalent roles,” and companies that hire Kenyan workers would have to disclose what they actually pay against them. Public comments close on 4 August.
Kenya is one of the world’s main hubs for this work, most of it done for foreign clients. Setting a benchmark against international rates puts the wage gap in writing.
The draft would also issue workplace protection guidelines covering written contracts, mental-health support, grievance procedures and working conditions. Annotation and moderation workers, the policy says, face exposure to harmful content and insecure conditions that existing rules do not adequately address. It says the rules would apply to local and international operators alike.
What the draft does not do is set a wage floor. It proposes benchmarks and disclosure. The numbers themselves would come later, in a framework still to be written.
Two things make that harder. The work falls to the Ministry of Labour and Social Protection, not the ICT ministry that wrote the policy — and the budget is thin. The government has costed the protection guidelines at KSh 50 million over three years, the pay framework at KSh 30 million, and a scheme to convert annotation experience into formal qualifications at another KSh 30 million. Together that is under $1 million, part of it expected from donors.
The draft also proposes a National AI and Other Emerging Technologies Council to coordinate policy, created by a new law the government wants drafted within six months and passed within twelve.
But parliament is already considering a different one. The Artificial Intelligence Bill 2026, from Senator Karen Nyamu, was gazetted on 19 February and had its first reading on 2 April. It would create an independent Office of the Artificial Intelligence Commissioner — a single regulator on the European model, insulated from the ministry.
So Kenya has two blueprints for governing AI, and they do not match. The 226-page policy never mentions the bill. The bill says nothing about annotation or moderation work; its only labour clause covers job losses caused by automation.
For now, the protections exist only in the document that carries no legal force.
