Kenya Union of Gig Workers- KUGWO

Success for me is not a long document that sits on a shelf. The first thing I would look for is a clear agreement that no AI policy process at any level can move forward without workers at the table. I don't meant workers at the table as guests or survey respondents whose experiences get summarised into statistics and someone else presents. Workers as decision-makers with a direct voice in what gets agreed. Second, I want to see agreement that when a platform uses an algorithm to block a worker's account, that counts as dismissal. It should trigger the same rights any fired worker has: an explanation, a chance to respond, a way to appeal. Right now a person can lose their entire income because a system flagged them overnight and no human ever reviewed it. That is not acceptable, and any dialogue that does not address it has missed the point. Third, the countries producing the most AI labour, Kenya included, should leave Geneva with a concrete plan to fund skills development for workers whose jobs are being automated. Again, this doesn't meant platform PR partnerships, I am talking about actual public investment, designed together with unions and delivered to workers. Finally, I want the dialogue to produce an honest conversation about who owns the data workers generate. Millions of people in the Global South have spent years labelling images, moderating content, and training systems they will never benefit from. That contribution needs to be named, counted, and eventually compensated. If the dialogue produces those four things: real inclusion, dismissal rights, funded reskilling, and data credit for workers, I will call it a success. If it produces a statement of good intentions, I will not count that as a success.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building

KUGWO represents workers who interact with AI every day as a condition of their employment. Our four priorities are not positions come from our lived experiences as gig and platform workers. The rights question comes first because everything else depends on it. Platforms classify workers as independent contractors to avoid all obligations tied to employer-employee relationship. Until that evasion is closed, every other protection is optional for them and unenforceable for us. Transparency is where that fight becomes concrete. A worker wakes up to a blocked account, they never get any explanation or a chance to appeal and be heard by a human being. We are not asking for a perfect system. We are asking for one that can be questioned. The social and economic implications cluster matters because the damage is already here. Task rates are dropping and work is disappearing. The people absorbing this are the same ones with no savings, no safety net, and no time to wait for a policy cycle to catch up. On capacity building, Kenya supplies significant AI labour to the world. Our members trained models they will never benefit from. KUGWO negotiated our own upskilling partnership with Microsoft because no institution was moving. We are not talking about this just as a success story, we want to point this out because that is a gap that governance should have filled. These four connect but more than that, they are the minimum. If the dialogue cannot deliver on these, it has not yet reached the people who need it most.

Liability is missing from every theme on the list. When a system blocks an account wrongly, cuts pay below what a person can live on, or puts a content moderator through material that damages them, nobody answers for it. The platform points to the algorithm and we all know that algorithm is not a person. Work on platforms crosses borders but consequences do not. The companies coordinate globally to write rules, set prices and take action against workers. A worker in Nairobi who wants to challenge a decision faces a legal system that was never designed for work with no fixed location. By the time any process concludes, the worker has already lost. Gender is everywhere in this conversation but named nowhere in the themes. The lowest paid categories on every platform are the ones filled mostly by women. Content moderation, which carries the heaviest personal cost, lands hardest on women in countries like Kenya. When training opportunities open up, men reach them first. A framework that does not ask who is actually being protected, and track the answer by gender, will write rules that sound fair and work out differently in practice. We need to remember that the people most harmed by how platforms operate are also the most invisible to the people writing the rules. Let's ensure that that doesn't continue at the global level

Kenya has many platform workers doing the work that feeds AI systems globally. That scale should give workers power but it has not, because nothing in the current setup converts numbers into rights. The most immediate challenge is that the governance gaps we selected do not exist in isolation here. A worker with no legal recognition cannot invoke a transparency right. Parliament ignored submitted memorandum's sent by gig workers organizations and granted global technology companies immunity from accountability in Kenya. The opportunity is specific. Kenya is actively positioning itself as an AI investment destination. That gives the government a reason to care about worker protection that goes beyond fairness. Investors and companies need a stable, skilled workforce. A workforce with no income predictability, no appeals process, and no skills pathway is a risk to that ambition. East Africa's platform workforce is mostly people entering employment for the first time. The rules being written now are the only rules they will know and we run a risk of having them believe that the world of work as it is right now is the norm.

Every other body setting AI standards right now represents either the countries building AI systems or the companies profiting from them. The AI Dialogue is the first space where every country has an equal seat regardless of whether it is a technology producer or a technology market. That is the most important thing it can do with international cooperation: use that universal membership to make sure the rules reflect the world, not just the wealthiest part of it. The second thing it can do is close the cross-border escape route. Right now a platform can face accountability in one country and restructure through a subsidiary in another. National law stops at borders. Platform operations do not. Meaningful cooperation means governments agreeing on a shared mechanism that follows the platform wherever it operates, not just where it chooses to be registered. The third is about timing and connection. The ILO finalises its binding Convention on decent work in platform work in Geneva in June 2026. The AI Dialogue opens in Geneva in July 2026. The ILO will set labour standards but it cannot govern the automated systems enforcing those standards on workers daily. The AI Dialogue can. If these two processes leave Geneva having never spoken to each other, platforms will use the gap between them to avoid accountability in both. International cooperation here means something very specific: two bodies in the same city in the same month deciding to build a bridge rather than two separate monuments.

Three processes are already doing important work that the Dialogue should connect with rather than duplicate. The African Union endorsed its Continental AI Strategy in July 2024. It covers governance, capacity building and ethics across 55 member states. The problem is it has no binding obligations, and by late 2025 over 80% of AI investment on the continent was concentrated in just four countries. A strategy without enforcement and without reach is not a solution. The Global Digital Compact, adopted in 2024, already commits governments to protecting workers from arbitrary algorithmic decisions and applying labour rights regardless of how work is done. Those words are in the document. What is missing is any mechanism to make them real. The Dialogue is the space where that commitment either gets built into something enforceable or quietly disappears. The ILO finalises its binding Convention on decent work in platform work in Geneva in June 2026, the month before the Dialogue opens. That Convention will set labour standards. It will not govern the systems that apply those standards to workers every day. That gap sits exactly where the Dialogue operates. The added value the Dialogue brings is not another framework. It is connection. The AU Strategy talks to African governments. The ILO Convention talks to labour ministries. The Global Digital Compact talks to technology ministries. None of them talk to each other. The Dialogue's specific contribution is to sit at the intersection of all three and make sure the protections each one promises actually land on the people who need them.

The Dialogue will reflect who it actually lets in, not who it says it welcomes. This is because the format and structure will not be procedural details but it will decide whose experience shapes the outcome. On stakeholder contribution, the most useful thing different groups can bring is what they cannot get from each other. Governments bring regulatory authority, companies bring technical detail and workers bring the evidence of what these systems actually do when they reach a person's life. That last contribution is the one most consistently missing from governance processes, and the one that would most change what gets decided. Worker organisations should have a dedicated track with the same standing as employer and government delegations, structured the way the ILO organises its tripartite process otherwise their representation will end up looking like tokenism. On format, three things would make the July 2026 session genuinely different from what has come before. Regional preparatory meetings before Geneva. The conversation should not start when delegates land in Switzerland. Organisations like KUGWO need space to consolidate worker evidence, align positions and arrive with something to contribute rather than just observe. Those meetings need funded interpretation and coordination support and trackable outputs. Every submission made to this Dialogue, including this one, should receive a written response indicating whether and how it was considered. A dedicated session connecting the outcomes of the June 2026 ILO Convention negotiations to the July Dialogue agenda. The two processes are in the same city in consecutive months and they should not be treated as if they are unrelated. The Dialogue's credibility will be measured by whether the people most affected by AI can point to a specific decision it changed.

The people least visible in AI governance are often the ones closest to how these systems actually work. Workers managed entirely by algorithms are the most direct example. Ride hailing drivers, delivery riders, microworkers whose pay, ratings and account status are decided by systems with no human behind them. These workers are the subject of almost every governance conversation but almost never present in one. The Dialogue needs asynchronous submission options, mobile-first engagement tools and compensation for participation time the same way it compensates expert consultants. Content moderators cannot speak publicly about their conditions because platform contracts use non-disclosure agreements as a lid on exactly the evidence this Dialogue needs. Including them means either legal protection for workers who speak to governance processes or anonymised testimony mechanisms that protect identity while capturing experience. Data annotation workers built the training sets that current AI systems run on. They have no formal identity in any governance conversation, treated as inputs rather than participants. Giving them standing starts with recognising their contribution legally, not just rhetorically. Informal economy workers sit outside the platform system entirely but are being displaced by it. A translator whose work has been automated, a transcriptionist replaced by a tool trained partly on their own previous output. These people have no route into a process conducted in Geneva in six UN languages.

Most governance dialogues are built around statements. Someone speaks while another listens and then a document is produced. We believe these formats would make this one different. Reverse panels. Workers question platform representatives and government delegates directly. A driver asks why account deactivation has no appeal or a content moderator asks why NDA clauses are legal even if it means eroding the mental health. Usually, the conversation changes when the direction of questioning changes. Scenario testing. Every proposed rule gets put in front of a worker before it is finalised. If she cannot explain how it would have helped her last month, it goes back for revision. Worker blocs. Worker organisations consolidate their positions together and negotiate as a group, the same way governments do. From experience, individual submissions get ignored while collective ones are harder to dismiss. Open drafting. Outcome text is visible to everyone between sessions. Participants can see whether their evidence shaped the language or disappeared. Evidence sprints where one specific problem is given one half-day session and a clear draft response. This then provides proof that the Dialogue can move from evidence to action without a two year wait.

The examples that have actually changed something share one thing in common. They attached a consequence to bad practice because principles without consequences have not moved platforms. Policy that shifted the burden- The EU Platform Work Directive, adopted in 2024, reversed a single assumption that had protected platforms for years. Previously workers had to prove they were employees. Now platforms have to prove workers are not. It also requires that any automated system used to dismiss or discipline a worker must be explainable and open to human review. Kenya's Labour Laws Amendment Bill, currently under discussion, is moving in the same direction by introducing a dependent contractor category for workers who rely heavily on one platform for income. A practice worth replicating- When KUGWO submitted evidence to the ILO Platform Economy process, we collected the data ourselves directly from workers rather than relying on what platforms reported. That approach produced a different picture. Worker-collected evidence as a formal input into governance processes is something the Dialogue should adopt as a standard practice, not treat as an exception.