What happened
NEAR AI, the AI-focused arm of the NEAR Protocol ecosystem, integrated its private inference service into Corbits, an enterprise AI platform, CryptoBriefing reported Wednesday. The integration lets Corbits customers run inference on proprietary or regulated data with hardware-enforced confidentiality, meaning the underlying compute host - and by extension the platform operator - cannot see the raw prompts, embeddings, or outputs flowing through the model.
Under the hood, private inference of this kind typically relies on trusted execution environments such as Intel TDX or Nvidia's confidential-computing GPU modes, paired with remote attestation so the client can verify that the workload actually ran inside a sealed enclave. Neither party disclosed contract value, customer counts, or a token-side commercial arrangement in the announcement.
The deployment is live rather than a research preview, which matters: enterprise buyers treat pilot-stage confidential compute very differently from production-ready inference they can point auditors at.
Why it matters
Confidential compute has been the missing piece for enterprise AI adoption in regulated industries. Banks, hospitals, and law firms have models they want to run and data they legally cannot expose to a third-party inference provider. The standard workaround has been on-prem GPUs, which is expensive and slow to scale.
Private inference over attested hardware collapses that trade-off: the customer keeps cryptographic assurance their data stayed inside the enclave, while the operator handles capacity. For NEAR, this is a positioning win in a crowded field. Phala Network, Oasis, iExec, and Super Protocol are all chasing the same enterprise AI + confidential compute wedge, and most of them are still shipping demos rather than integrations into commercial platforms.