Google's choice of Apache 2.0 for Gemma 4 is a structural move against the walled-garden model that defines most frontier AI deployment. Apache 2.0 imposes no restriction on commercial redistribution — a developer can fine-tune Gemma 4, ship it in a product, and charge for that product without any revenue share or licensing negotiation with Google. That is a materially different arrangement than what most comparable releases permit, and it is not accidental. Google has maintained open-source commitments since 2005, but attaching that licensing posture to a model built on…