SGLang RCE exploited at NEAR AI; user and agent data taken
Original title: SGLang RCE Exploit Breaches NEAR AI, Steals Data From 18 Agents and About 100 Users
About 100 Private Chat accounts and 18 agents were exposed; two Cloud API keys were revoked and no CVE was named.
An attacker exploited a remote code execution flaw in SGLang on a NEAR AI inference server last Sunday and could view that day’s requests, including confidential-mode traffic inside a TEE. The attacker later reached deprecated Private Chat and an older agent deployment outside the TEE, downloading details for about 100 users and data from 18 agents, including two customers’ Cloud API keys. NEAR AI revoked those keys and sessions, patched SGLang on its servers, and restored NEAR AI Cloud. It reported the flaw upstream but gave no CVE, versions or upstream fix.
Unconfirmed: The CVE, affected SGLang versions, an upstream fix, and whether Sunday request users or Brave Leo chats were affected are unstated.
- Private Chat exposure was emails, names and, for some, plan and balance data.
- No stored Private Chat conversations or uploaded files were found accessed.
- Agent data included chat history, settings and logs.
- Users not emailed by security@near.ai were told they need do nothing.
Impact: RCE on an inference server and theft of account and agent data
| Product | Affected |
|---|---|
| SGLang | — |
- 1.Be wary of unsolicited messages about a NEAR AI account or balance.
- 2.Users contacted by security@near.ai should follow the steps sent to them.
- ATT&CK
- T1190 Exploit Public-Facing ApplicationT1005 Data from Local SystemT1552.001 Credentials In Files
- Victims
- NEAR AI
From the original article
2 min read at thehackernews.com
An attacker exploited a remote code execution flaw in SGLang, an open-source inference engine, and ran code on one of NEAR AI's inference servers last Sunday. The attacker could also have seen requests being processed on that server that day, including requests sent in confidential mode, the company said earlier this week. That mode processes requests inside a trusted execution environment (TEE
thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.