AI and Open Access: why libraries must reclaim the terms of “openness”

Purpose This opinion piece argues that the encounter between Open Access (OA) and generative artificial intelligence (AI) is not a peripheral concern but the decisive test of whether the OA movement remains committed to its founding principles in an era of large-scale machine reading. Design/methodology/approach The paper is a critical viewpoint, drawing on recent peer-reviewed literature (2025–2025), legal commentary, publisher policy development and library sector survey data to analyze three flashpoints where OA and generative AI collide: authorship and literature inflation, equity and the Global South and the economics of scholarly publishing. Findings Major publishers are quietly reenclosing OA content through terms-of-service overlays that purport to forbid AI training on CC VY licensed works. AI literacy programming, while necessary, is insufficient. Academic libraries as machine-readable infrastructure, refuse tiered openness in consortial negotiations and invest in shared, library-led AI tooling as public infrastructure rather than ceding the AI mediation layer to commercial vendors. Originality/value The piece reframes generative AI not as an external shock to OA but as a long-deferred stress test of its first principles. It moves the library response beyond literacy toward concrete licensing, contractual and infrastructural commitments, and warns that the next two to three years will determine whether OA survives the AI era or dissolves into a new enclosure dressed in machine-readable clothing.