This article offers a 2025 observation of overseas Scientific, Technological and Medical (STM) journal publishing and argues that large-scale application of Artificial Intelligence (AI) has pushed the sector into a technologically accelerated phase of paradigm transformation. In this phase, governance deepening becomes the year's central thread, cutting across the exploration of openness models, publishers' strategic transformation, innovation in recognition mechanisms, and the delineation of ethical boundaries. The paper therefore treats "AI acceleration" not as a narrow technological upgrade but as a governance stress test that forces the scholarly publishing system to rearticulate its normative foundations, role definitions, and power arrangements. The analysis starts from a double-edged reconfiguration. On one side, the deepening of open science and the upgrading of technical tools mutually empower the sector's development by enabling new modes of dissemination, workflow restructuring, and infrastructure modernization. On the other side, this same empowerment generates systemic disruption: traditional academic ethics encounter new pressures; the positioning of publishing entities is contested; and governance models that previously focused on surface-level compliance become insufficient when technology penetrates deeper into decision-making and evaluation. The paper's logical pivot is a shift in what the industry is "about." It argues that transformation has moved beyond iterative tool replacement and has become a contest over dominance in knowledge production and dissemination, with the sovereignty of knowledge production as the core issue. Correspondingly, the focus extends from the free flow of knowledge to the redistribution of power. To make this power-centric framing concrete, the article names several emblematic phenomena. It points to the APC dilemma in the open access movement, to the transformation of scholarly content into AI training assets, and to algorithmic involvement in academic judgment. These phenomena are framed not as isolated controversies but as expressions of complex tensions among technology, commerce, and academia. The article further argues that publishers' AI-enabled deep involvement in the entire research process increases the potential risk of privatization and black-boxing of academic infrastructure. In this framing, the central governance problem is not merely "openness versus closure, " but transparency versus opacity, public-interest orientation versus infrastructure enclosure, and accountable evaluation versus algorithmic intervention without clear boundaries. In response, the paper observes that overseas STM publishing communities are attempting three governance moves: pushing open science from nominal openness to substantive openness; promoting publishers' role transformation; and exploring academic ethics and evaluation systems that can adapt to the AI era. The article concludes that these moves collectively foreshadow a governance trajectory in which the sector may shift from localized adaptive adjustments toward more comprehensive systemic reconstruction. The overall contribution is a justice-oriented interpretation: openness must be evaluated not only by access outcomes but also by how governance designs distribute power, protect academic integrity, and prevent infrastructure opacity in an AI-accelerated publishing system.