Open weights are not open source: Why AI's favorite label is under dispute

The distinction between "open weights" and "open source" in AI is significant, as it determines the extent to which users can interact with a neural network. While open weights allow deployment, true open source grants the freedom to inspect, reproduce, alter, and redistribute the system that created the network. This difference is crucial for meaningful engagement with AI systems beyond mere deployment.

This report highlights a key definitional dispute, clarifying that "open weights" only allow deployment, unlike "open source" which permits full inspection, reproduction, alteration, and redistribution.