NEAR AI says it has contained unauthorized access and restored cloud services, with account information of about 100 users downloaded.

November 10 - According to The Block, NEAR AI stated in its latest security incident report that it has contained unauthorized access to some of its infrastructure and has restored NEAR AI Cloud services. The report states that attackers exploited a remote code execution vulnerability in the open-source inference engine SGLang running on its servers to execute code on an inference server and thereby gain access to related infrastructure. NEAR AI said it has shut down the relevant access, rotated affected credentials, and patched the vulnerability on all inference servers.

According to the report published on November 9, the investigation has so far confirmed that the attackers subsequently accessed internal systems behind the decommissioned Private Chat and downloaded account information of approximately 100 users, including email addresses, names, and plan and balance information for some accounts. In addition, 18 NEAR AI Agents running on legacy deployments and not in a Trusted Execution Environment (TEE) were accessed, with data taken including chat histories, settings, and logs; the API keys of two Cloud customers appeared in the logs, and the relevant keys have been revoked. The report states that on November 8, the attackers may also have observed user requests being processed on affected servers, including requests in confidential mode, but no evidence has been found that stored Private Chat conversations or uploaded files were accessed. NEAR AI previously stated that NEAR Protocol and other NEAR services were not affected by this incident.

NEAR AI said that after discovering the incident, it revoked all active sessions and administrative credentials, replaced all Agent Tokens, and took Private Chat offline, while completing the remaining credential rotation and a comprehensive forensic review. Affected users have been notified by security@near.ai, and the owners of the 18 relevant Agents have separately received explanations. The company has reported the vulnerability to the SGLang project and will make regulatory notifications as required.