Agencies and Ownership: How Corporate Backing Shapes Open-Source Growth on GitHub

Abstract

As agentic AI systems emerge, it remains unclear whether corporate involvement provides agentic repositories with a growth advantage relative to traditional AI projects. This study investigates whether corporate ownership predicts greater community adoption over time, whether this advantage is larger for agentic than non-agentic repositories, and whether the advantage widens or narrows as repositories mature. A dataset of 800 AI-related GitHub repositories was collected using the GitHub API, balanced across ownership (corporate vs. community) and project type (agentic vs. non-agentic) using keyword matching. Stars and forks were recorded at a single snapshot, with forks tracked across four checkpoints per repository (spanning 2015-2026), as measures of community adoption. Cross-sectional ordinary least squares (OLS), longitudinal mixed-effects panel, and weighted least squares (WLS) growth-rate regressions were applied to estimate differences in adoption across ownership and project-type. Corporate repositories start with lower adoption (beta = -4.70, p < 0.002) but grow faster with age (beta = 0.22, p < 0.001), eventually overtaking community projects. This advantage is larger for agentic repositories in cross-sectional and growth-rate models (beta = 4.19, p < 0.013), though not in the panel models with repository-level random intercepts controlled (p = 0.159). The corporate advantage widens over time for non-agentic repositories but attenuates for agentic ones (beta = -0.85, p = 0.003). Our results imply that corporations do not immediately dominate open-source agentic artificial intelligence, leaving space for independent developers to remain competitive.