A new arXiv paper argues that today's mobile networks are poorly equipped to support edge AI inference. While 5G systems can transport tensors—the data structures AI models consume and produce—they do not expose or understand the inference context that gives those tensors meaning. That gap makes it hard to deliver inference as a reliable, manageable network service.

The authors propose exposing AI inference as a telco network API over the 5G user plane. This would let the network recognize and handle inference requests like other service-level operations, rather than treating them as opaque data transfers. The paper is positioned as a step toward integrating AI capabilities into the mobile network's native service model.

Because this is a single source, there are no conflicting findings to compare. The abstract is brief, so the article sticks to the core proposal: making inference context visible to the network and turning it into an API-level service.