From Invisible Labor to Worker-Led AI Accountability in Kenya

A project reflection by the Equiano Institute, supported by the Pulitzer Center through the South to South (S2S) AI Accountability CoLab microgrant.

The Fair Work for AI Kenya project began from a simple premise: The future of artificial intelligence in Africa cannot be understood only through models, platforms, investment announcements, or national strategies. It must also be understood through the people whose labor makes AI possible. In Kenya, data workers, content moderators, annotators, red-teamers, and student contributors are part of the human infrastructure behind AI systems. Yet their experiences are often missing from policy debates, university conversations, and public narratives about technological progress.

Where the project sits in the AI accountability landscape 

Kenya has become a significant hub for AI data work and digital labor. This creates opportunity, but it also creates accountability risks. Workers may face opaque recruitment, low or inconsistent pay, psychologically difficult content, strict confidentiality rules, limited grievance channels, and few ways to influence the systems they help build. At the same time, global AI companies, platform clients, vendors, and policymakers often speak about innovation without fully accounting for the labor conditions beneath it.

Fair Work for AI Kenya sits in this gap. It connects AI accountability to labor rights, data dignity, privacy, wellbeing, benefit sharing, and institutional responsibility. The project is relevant to holding powerful actors accountable because it shifts attention from abstract promises about AI to concrete questions: Who performs the work, under what conditions, who profits, who is protected, and what obligations should platforms, clients, universities, and regulators carry?

How the grant and network strengthened the project 

The Pulitzer Center microgrant gave the project the implementation capacity to move from concept to practice. It supported coordination, workshop preparation, facilitation, participant access, materials, documentation, design, website development, and final outputs. Just as importantly, the South to South AI Accountability frame gave the project a clear public-interest orientation: Journalism, civil society, and academia can work together to make AI accountability tangible for communities affected by AI systems. (Learn more about our workshop here and our session at London Data Week here) 

The network also helped us position the work beyond a single event. Fair Work for AI Kenya became a bridge between worker experiences, student learning, civil society evidence, and policy-facing discussion. It gave us a foundation for future work: a reusable teach-out model, an open resource page, and a possible next phase focused on worker-led accountability tools.

How journalism informed the work 

Accountability journalism was central to the project. It gave us credible, accessible case studies on AI labor, AI colonialism, platform accountability, and Global South AI supply chains. This journalism helped participants see that AI accountability is not only a technical problem. It is also a labor problem, a governance problem, a rights problem, and a narrative problem.

The reporting was especially useful because it translated complex systems into human stories. That made it easier to engage students, civil society actors, and policy stakeholders who may not have a technical background. We used reporting as a starting point for discussion: What happened, who was affected, what accountability failed, and what would need to change?

The main limitation is that journalism can open the door, but it cannot substitute for local organizing, worker testimony, legal analysis, or sustained policy engagement. Reporting can reveal patterns and make harms legible, but communities still need tools, institutions, and resources to turn that knowledge into action. Our next phase is designed to address that gap. 

Challenges during implementation 

The broader political and economic climate made the work harder. AI hype has encouraged governments, companies, and universities to focus on competitiveness, productivity, and attracting investment. Those goals matter, but they can crowd out questions about labor conditions, psychological safety, consent, data dignity, and fair compensation. In that environment, worker-centered accountability can be treated as secondary, even though it is foundational to responsible AI development.

Implementation also required careful coordination. We had to translate journalism into workshop materials, develop accessible language for non-technical audiences, coordinate partners, prepare documentation, and handle sensitive labor issues responsibly. We responded by keeping the framing practical: Workers are not just inputs to AI systems. They are experts whose judgments shape what AI systems see, learn, filter, and amplify.

 Workshop highlights 

1. “Accountability means harm cannot be outsourced.” said Nomusa Nkwanyana, researcher at Zibula Advisory.
   If a company benefits from labor, it must remain responsible for the conditions of that labor. Sonia Kgomo, organizer at African Tech Workers Rising described recruitment under an admin advert, late disclosure of Nairobi content moderation, NDAs used as silencing tools, leave denied for nearly eighteen months—then dismissal after organizing, which led to African Tech Workers Rising.
2. “There is no clear legal classification for this category of digital work.” Without a baseline for wage, leave, and rights, accountability and compensation have nowhere to stand. This calls for a legislative baseline for content moderators and data workers. 
3. “The fight over digital labor in Africa is the fight for young Africans’ future of work.” said by Odanga Madung, a data journalist and researcher who spoke at the workshop. 
   “Digital jobs” remain a political promise even as conditions stay precarious and poorly governed. “This is the promise that our political class has given our young people. The fact that we are losing, and that this thing is expanding while the voice of people who care about this problem cannot get the attention it used to, is very concerning.” said Madung.
4. The workshop closed with demands for plural evidence, human-rights-centered design, and shared narratives that can bridge technologists, policymakers, civil society, and affected workers. 

What comes next 

The next step is to turn Fair Work for AI Kenya into a worker-led accountability and education toolkit. This would include a polished teach-out package on AI labor and data dignity, a worker-facing guide on risks and rights in AI data work, a follow-up roundtable with workers and policy stakeholders, and a short policy memo for universities, civil society organizations, platforms, and regulators. 

Journalism will remain embedded in this next phase. Pulitzer Center reporting can continue to serve as the evidence base for case studies, workshop prompts, and accountability mapping. The goal is to make AI accountability usable: not only something people read about, but something communities can teach, discuss, adapt, and use to demand better practice.

If Africa’s AI future is to be fair, the people behind the data must be visible in the governance of AI. Fair Work for AI Kenya is one step toward that future: a future where AI workers are recognized not as hidden inputs, but as co-creators whose dignity, safety, and rights shape the systems the world will depend on.