Concurrent Session 2: Power in the Loop

Format: Reverse Panel 1

Moderator: Madelynn Martiniere

Speakers: Wayne Wang, Shivam Shukla, Anyamitra “Yanagi” Wongprasitthiporn, Patrick Feuerstein, Mohammad Shan

This panel asks when human oversight gives workers and affected communities real authority over AI, including the power to refuse a harmful deployment, rather than a formal role in approving an algorithm’s decision.

Shivam Shukla argues that the Solidarity Stack needs an accountability layer: institutions able to refuse harmful AI deployments and make that refusal binding. He points to statutory professional bodies and worker-led organizing as possible sources of this authority, asking how refusal capacity can be activated across professions and jurisdictions rather than leaving workers with only advisory roles.

Wayne Wang examines the often-invisible labor and data annotation that sustain systems marketed as “creative” AI. He asks who produces that creativity and who governs the supply chains on which it depends. Drawing on his work in AI governance and the Digital Majority World, he considers how law and institutional design can render these labor relations visible, accountable, and subject to meaningful human oversight, even amid intensifying global competition in AI.

Anyamitra “Yanagi” Wongprasitthiporn examines how community mapping and open source infrastructure can give local communities greater control over their digital environments. Drawing on TechTransThai’s work with OpenStreetMap and Linux, she asks how shared technical resources are built, maintained, and governed.

The workers who label data, moderate content, and check model outputs carry much of the AI industry, but the companies that buy their labor rarely see them. Patrick Feuerstein presents Fairwork’s new certification for AI supply chains. Fairwork assesses the firms that commission this data work against five principles: fair pay, fair conditions, fair contracts, fair management, and fair representation. Assessors combine company records with anonymous interviews with workers. Firms that take part commit to a plan for improvement and face follow-up checks. Feuerstein asks what certification can change for data workers in the Global South, and where it falls short of the power that worker ownership would bring.

Mohammad Shan examines how AI-based visual inspection in Malaysian electronics manufacturing draws on workers’ tacit expertise while leaving ownership and control with manufacturers and equipment vendors. His ongoing study combines analysis of job postings with interviews with inspectors, technicians, and quality engineers to investigate deskilling and changing roles. He proposes compensating defect labelling as data labour, giving workers a voice in setting AI reject thresholds, and co-designing portable reskilling pathways with worker organisations.

Community notes

Reporter 1: Anyamitra “Yanagi” Wongprasitthiporn
Reporter 2: Patrick Feuerstein

Back to program