DATA Network Moves AI Training Registry to Sui

Sui Foundation and The DATA Foundation are bringing the DATA Network and its Trace AI-data registry into the Sui technology stack, with Walrus designated as the default storage layer for DATA Foundation applications. The partnership is intended to combine rights-cleared AI training data, provenance records and access-controlled storage inside the same infrastructure stack. Sui Foundation is also investing $30 million in the DATA Network as part of the long-term alignment.

According to the official Sui Foundation announcement, Trace has already registered more than 250 million data contributions and processes over 3 million receipts per day on average. Applications including Kled, Numo and Oto use the system to source data for AI and robotics customers. Those figures measure Trace registrations and receipts on the existing network, not data already migrated onto Sui or stored in Walrus.

Walrus Adds Storage to Trace’s Provenance Layer

Trace currently records where a data contribution originated, when it was registered and which rights accompany it. Under the planned Sui integration, each dataset and license can become a programmable object, while Walrus can store an encrypted, immutable reference copy accessible to authorized buyers. That would extend Trace from an audit trail toward a system where provenance records can be checked against the underlying reference data itself.

Walrus already operates as Sui’s decentralized data platform, with integrity proofs anchored to Sui. Earlier infrastructure development positioned Walrus around data markets and access-controlled AI storage. The DATA integration gives that storage layer a specific training-data workload, although the announcement does not establish how much data will initially migrate, how much storage will be consumed or what volume AI labs will purchase through the upgraded network.

Regulatory provenance is becoming more relevant as AI training expands. The EU AI Act requires providers of general-purpose AI models to maintain copyright-compliance policies and publish sufficiently detailed summaries of training content, while technical documentation can include information about training-data types and provenance. Trace may help organizations document data origin and usage rights, but using the network does not itself establish compliance with the AI Act.

Native Sui Migration Remains Ahead

The DATA Network will continue running in its existing form until the native Sui upgrade is complete. A detailed migration schedule and holder transition plan remain pending. The announcement therefore establishes a future protocol migration rather than an already completed deployment of the DATA Network on Sui.

Sui’s broader AI infrastructure has already been tested through programmable tunnels that reached 6.08 million interactions per second in a July experiment involving AI agents, payments and games. Those interactions occurred inside off-chain payment and state channels that settled back to Sui mainnet, so the benchmark demonstrated specialized tunnel capacity rather than 6 million ordinary mainnet transactions per second.

The next phase could extend beyond storage and provenance. DATA Foundation CEO Andrea Muttoni said the organizations may eventually explore verifiable trusted execution environments for model evaluations and fine-tuning. That remains a future development rather than an operational capability of the current integration. For now, the measurable transition is narrower: Sui is funding the DATA Network, Walrus is becoming its planned default data layer, and Trace’s existing registry is preparing to migrate into the Sui stack.