Table of Contents
UR, the private networking project on Bittensor Subnet 25, has added Persistent AI as a network operator. Persistent AI will route traffic from its Private Inference service through UR's distributed connections, adding IP-address privacy to a product built around private AI access.
Private Inference provides access to models such as GPT and Claude without storing users' chats and allows private payments. The integration extends those protections to the network layer. UR encrypts traffic and routes it through connections shared by people globally, keeping users' originating IP addresses private.
Chat history, payment identity, and network addresses expose different kinds of information. An IP address can reveal a connection's approximate location and internet provider even when a service does not retain conversations. UR's role is to shield that connection information rather than replace Private Inference's chat and payment protections.
Private Inference is built for both people and AI agents, with access to multiple models through one API key. Its private payment system uses stablecoins through Aztec and the x402 payment protocol.
The partnership also brings another customer-facing service into SN25's connectivity market. Network operators develop products and bring traffic, while providers supply the underlying connections. UR says operators pay for the traffic they use in SN25 alpha, the subnet's token, linking usage of services such as Private Inference to demand for that asset.
UR recently launched SN25 mainnet and introduced a Network Capacity Reserve to fund connectivity ahead of customer demand. The reserve is intended to keep provider capacity available as operators bring services onto the network.
Network integration is the teams' first joint project. Persistent AI also plans joint privacy research into zkAPI and private, open-model routing, along with an eventual Private Inference subscription offering for UR users.
