Selective transparency in open-source AI is creating a false sense of openness. Many companies, like Meta, release only partial model details while branding their AI as open-source. This article dives into the risks of such practices, including erosion of trust, ethical lapses, and hindered innovation. Examples like LAION 5B and Meta’s Llama 3 show why true openness — including training data and configuration — is essential for responsible, collaborative AI development.