(True) Private Inference

Not all inference is created equal. Some inference providers have been caught sandbagging serving LLM models to save costs on inference and while others are suspected but never fully caught. Furthermore, it is not fully transparent what is being done with the data stored by AI apps and labs. Some have been accused on training the data. In most cases, this is not desirable. Conversations between agents are to be private for personal and also for compliance reasons as in the case of doctors and lawyers.

Sedona attempts to solve this with true private inference. Any model that wins a competition on Sedona Marketplace, Sedona will provide an API endpoint to the model that is truly private.