AI Runs on Human Labour — But at What Price? - 2026

Although AI seems to be capable of doing everything “by itself”, in practice it depends on tens of millions of workers who mainly perform their work in precarious conditions. They annotate, train, correct and sometimes also impersonate today’s AI systems. They do this via online platforms such as Sama and Amazon Mechanical Turk, or at the “office” via Business Process Organisations. In a market where AI companies have astronomical valuations, they often work underpaid, with minimal rights and security, poor working conditions and a lack of psychological support. They are the hidden, but indispensable workforce behind AI. This session, co-organised by the WageIndicator Foundation and the Data Labelers Association, focuses on the question “what should a data worker earn at a minimum to be able to cover their basic needs?”. We bring together worker representatives and gig economy experts to debate this question. Central in this discussion is the Living Wage / Living Tariff methodology: a structured approach for setting a base pay on the basis of cost of living. Rather than presenting it as a finished product, the discussion will explore how this methodology might work in practice for outsourced AI work, where its boundaries lie, and who should take responsibility. Short polls will gauge participants’ views, while audience questions and reflections in the chat will guide the flow. Speakers will respond directly to inputs, creating an exchange that bridges lived experience with global perspectives. Key insights will be summarised and shared afterwards. We will conclude with reflections on how the cost-of-living methodology, combined with worker voice, could become a cornerstone for defining base wages and tariffs in AI-supply chains.

Note: This session was originally planned to take place at RightsCon. In solidarity with the RightsCon community and to continue the conversation on this important topic, we are hosting the event independently.