New study reveals 4 core tensions in defining open source AI

A new paper titled "Reimagining Open Source and Openness in AI" by Genevieve Smith and 22 co-authors, to appear at the 2026 ACM Conference on Fairness, Accountability, and Transparency (FAccT '26), presents empirical findings from a multi-sectoral workshop grounded in futures thinking and participatory design. The workshop brought together diverse stakeholders—beyond typical industry voices—to co-create desirable visions for AI's role in society and generate action pathways for responsible openness. The study challenges dominant corporate narratives by surfacing plural perspectives on what openness means in practice.

The paper's key contribution is identifying four core tensions that emerged when participants translated high-level aspirations into concrete actions: whether openness is an end or a means (purpose), whether it should focus on expanding access or ensuring meaningful equity (scope), whether requirements should be mandatory or conditional (operation), and whether openness alone is sufficient or must be paired with robust governance (dependency). These tensions reveal that responsible openness is not a singular technical solution but a negotiated sociotechnical project shaped by values, power, and positionalities. The authors also advance AI governance methodology by demonstrating how participatory futures methods can yield actionable research priorities beyond industry-led frameworks.