The research proposes a self-governed prompt commoning system for data workers and other collectives affected by fragmented, outsourced and algorithmically managed labor. I pursue the goal to respond to the opacity of AI chains, where microtasks, intermediaries and requesters are rarely public. I seek a way to transform workers' dispersed experiences into a formatted common archive of prompts which stores evidence with the power to transform workers' situated knowledge into self-governed domestic LLMs. These models can become the cornerstone for the creation of tools, critical artifacts or actions to expose precarious labor conditions as well as companies and build accountability across AI production chains.

Through qualitative interviews, situated ethnography, analysis of real microtasks and prompt commoning workshops, the research will explore how workers can collectively create, review, version and reuse prompts to document their labor traces. 

The expected contribution is a tested methodology and governance model for ethical context transfer. From labor traces to prompt archives, from prompt archives to local models, and from local models to situated interventions defined by the collectives. Its value lies in offering workers, unions, researchers, journalists and regulators a practical way to make AI labor and value chains more traceable and actionable.