NEAR AI has disclosed a security breach after an attacker exploited a vulnerability in the open-source SGLang inference engine running on its servers, exposing account details belonging to about 100 users.

The company published its security incident report on Monday, 9 November, and said it had contained the breach, fixed the flaw and restored access to NEAR AI Cloud. NEAR AI is the artificial intelligence arm of the NEAR ecosystem, and it runs AI inference on servers that process user requests.

According to the report, the attacker exploited a remote code execution vulnerability in SGLang, which allowed the unauthorised party to run code on one inference server. From there, the attacker reached an internal system behind Private Chat, a feature NEAR AI said had already been deprecated, and downloaded account details for approximately 100 users.

The downloaded details included email addresses and names, and for some accounts, plan and balance information. The attacker also accessed 18 agents on a legacy deployment of NEAR AI Agents that was not running in the company's confidential computing environment. That access covered the agents' chat history, settings and logs.

Requests on the server could have been observed

NEAR AI said the attacker could also have observed user requests being processed on the compromised server on 8 November, including requests running in confidential mode. The company said it found no indication that stored Private Chat conversations or uploaded files were accessed.

Two customers had Cloud API keys exposed through logs, and NEAR AI said it had revoked both credentials. Affected users have been contacted directly from security@near.ai with the steps that apply to them, and the company said users who have not heard from it do not need to take action.

On detection, NEAR AI said it revoked all active sessions and administrative credentials, replaced all agent tokens, took Private Chat offline and deployed a fix for the SGLang vulnerability across all of its inference servers. Access to NEAR AI Cloud, which was paused during the investigation, has been restored.

NEAR Protocol blockchain was not affected

NEAR AI said the incident did not affect the underlying NEAR Protocol blockchain or other services in the NEAR ecosystem. It also said it had notified the SGLang project of the vulnerability.

The breach is a reminder that flaws in the software powering AI services can expose user data even when a service uses environments designed to protect it. NEAR AI's confidential computing setup is meant to shield data from the operator and the host, but the company said software running inside that environment can still be vulnerable to flaws.

Read NEAR AI's security incident report and the report on the breach by The Block.