NEAR AI Halts Cloud Inference After Detecting Unauthorized Access

NEAR AI suspended its cloud inference service on November 9, 2026, after detecting unauthorized access to its internal systems, according to the company's disclosure. The confidential-AI provider says customer prompts and workloads remained protected inside hardware-isolated Trusted Execution Environments, and that NEAR Protocol was unaffected.

The company suspended NEAR AI Cloud inference operations immediately after spotting the breach. By November 10, 2026, it confirmed the unauthorized access had been addressed and closed, but the cloud service stayed offline until it could be safely restored. An investigation into the incident is still continuing.

What is NEAR AI's defense?

NEAR AI told users that customer workloads kept running inside hardware-isolated Trusted Execution Environments (TEEs), specifically pointing to Intel TDX and NVIDIA confidential computing as the isolation technologies. That architecture is the core of its reassurance: the company says the setup keeps data confidential not only from the cloud provider but from NEAR AI itself.

The breach landed at an awkward moment. NEAR AI launched a cybersecurity-focused initiative on October 29, 2026, offering uncensored models for security research under the same TEE-based privacy standards it applies elsewhere. Less than two weeks later, the vendor was disclosing an intrusion of its own.

What remains unknown?

Plenty. NEAR AI has not described how the access occurred, who was behind it, or which specific systems were touched beyond calling them internal. Little else about the incident has been made public so far.

The interesting wrinkle here is the gap between a company getting breached and customer data getting exposed — and NEAR AI is arguing those are not the same thing. If the enclave design held up as the company says it did, this incident becomes a live test of the confidential computing model: an intruder inside corporate systems would, in theory, remain locked out of prompts and workloads running in the TEEs.

That is a meaningful claim, but it is the company's own. The investigation remains open, and no independent confirmation of what the intruder could or could not see has been published.

What does this mean for the NEAR ecosystem?

For a firm whose entire sales pitch is confidential AI, the reputational stakes are high. Analysts tracking the incident note that investors may view the breach as a question mark over platforms handling sensitive data, and any fallout could dent NEAR AI's standing or spill over to the broader NEAR ecosystem.

The company has tried to wall that risk off. It stated plainly that NEAR Protocol and its other NEAR services were unaffected — the blockchain side of the house kept running while the AI side locked its doors.

There is also a regulatory dimension worth watching. The same analysis flags potential long-term regulatory scrutiny of the tech sector as the investigation plays out, particularly for vendors marketing privacy guarantees built on enclave-based architectures.

The cloud inference service remains suspended until NEAR AI deems it safe to restart, and the company has not set a public timeline for either the service's return or the close of its investigation.