NEAR AI Restores Cloud Service After SGLang Flaw Exposes User Data
NEAR AI restored its Cloud service after an attacker exploited a remote code execution flaw in the open-source SGLang inference engine on November 8. The company says about 100 users' account details and 18 legacy AI agents were accessed, while the NEAR blockchain itself was not affected.
Written by CoinArticle’s AI Newsroom · from 2 cited sources. How we work
What Was Exposed
NEAR AI said an attacker exploited a remote code execution vulnerability in SGLang, an open-source inference engine it runs on its servers, on November 8. Its security report, published November 9, put the affected scope at approximately 100 users and 18 AI agents. The company said NEAR Protocol and other NEAR services were not affected.
For the roughly 100 users, the exposed records included names, email addresses, and in some cases subscription plan and account balance details. NEAR AI reported no evidence that stored Private Chat conversations or uploaded files were accessed. The attacker reached an internal system tied to Private Chat, a service the company has since discontinued.
Agents and Credentials
The 18 agents ran on an older version of NEAR AI Agents that was not inside the company's trusted execution environment, the hardware-based isolation layer used for confidential computing. Information retrieved from those agents included chat histories, configuration settings and operational logs. Two NEAR AI Cloud customers had API keys appear in those logs, and the company revoked both.
NEAR AI also said the attacker could have observed requests processed on the compromised inference server on November 8, including some requests running in confidential mode. The report did not say how many requests were visible, and the company has not stated whether any were read.
Restoration and Response
NEAR AI suspended Cloud inference after detecting the intrusion, closed the attacker's access, and kept operations paused while it completed remediation. The Cloud service has since returned online. The response included revoking active sessions and administrative credentials, replacing agent tokens, taking Private Chat offline, and applying the SGLang fix across all inference servers.
The company said it completed a forensic review and contacted users whose accounts or agents were confirmed accessed. Its public update states that users who have not heard from the company need take no action. It advised caution with unsolicited messages referencing account information or balances, and said regulatory notifications would follow where required. Crypto.news said it reached out to NEAR AI for comment and received no response by publication.
Reading the Scope
Coverage of the incident differs on emphasis. Crypto.news framed the exposure around confirmed access and the company's containment steps. NewsBTC stressed that the counts describe exposure rather than losses: 100 user records and 18 agents do not establish that 100 users lost funds or that every request was read. Neither outlet reported fund losses, and the company's account does not describe any.
NewsBTC also noted that the incident concerns AI-serving infrastructure, not the NEAR blockchain's consensus rules or NEAR wallet keys in general. Users who used credentials with related services should review permissions and rotate keys, according to that analysis. The completeness of the remediation remains a question the company's disclosures do not fully settle, since the reports do not establish whether other access paths existed.
Why It Matters
For Traders
No NEAR Protocol or wallet compromise was reported, so near-term price relevance is limited, though watch for any follow-up disclosing fund losses.
For Investors
The incident highlights that offchain AI infrastructure can expose user data even when the underlying blockchain is unaffected, a risk to weigh for AI-integrated crypto projects.
For Builders
Legacy agent deployments outside trusted execution environments and logs holding API keys were the weak points, so audit inference dependencies and log redaction now.
This article is for information only and is not financial advice. Read the full disclaimer.
