NEAR AI says an attacker used a flaw in SGLang to reach its systems: about 100 users, 18 old agents and 2 API keys exposed
NEAR AI reports that on 8 Nov 2026 an unauthorized party exploited a remote code execution flaw in SGLang, the open-source engine that runs its models, and reached an internal system. About 100 users' account details, 18 legacy agents' data and two customers' API keys were exposed. What the company says, what is still unknown and what to check.
Updated 10 Nov 2026: see what changedHow urgent
Plan this week
Who is affected
NEAR AI users and agent owners, above all those the company emailed directly, and teams that run SGLang themselves. NEAR AI says about 100 users, 18 agents on a legacy deployment and two Cloud customers' API keys were affected. We do not know which SGLang versions have the flaw or whether other deployments were attacked
What to do
If NEAR AI emailed you, follow the steps in that email and replace any API key you used there; if you run SGLang yourself, look for the project's advisory (we could not find one) and keep the server off the public internet; if neither applies, NEAR AI says no action is needed
Free checklist
Published 10 Nov 2026 · Sources last checked 10 Nov 2026No. 70 / 87
Updates, follow-ups and corrections
- Update · 10 Nov 2026. First version of the story, based on NEAR AI's own incident report (read through a subagent) and two press reports. We found no CVE, affected versions or fixed version for SGLang in any source. The company's description of what was exposed is its own; nobody independent has confirmed it.
We keep the original text. When something changes, when we follow a story, or when we get something wrong, we add a dated note here instead of editing silently.
What happened
NEAR AI says that on 8 November 2026 an unauthorized party exploited a remote code executionRCE (remote code execution) A flaw that lets an attacker run their own code on your server or device from a distance. Tap again to open the glossary flaw in SGLang, the open-source inference engine that runs on its servers. That gave the attacker code execution on one inference server and then access to infrastructure, including an internal system behind a deprecated product called Private Chat. The company published its incident report on 9 November, paused its Cloud service during the investigation and says the service is back.
What NEAR AI says was exposed
This is the company's own account; we have not seen an independent confirmation.
- Users: account details (email address, name and, for some accounts, plan and balance information) for about 100 users.
- Agents: data from 18 agents on a legacy NEAR AI Agents deployment that did not run inside its protected environment: chat history, settings and logs.
- API keys: two Cloud customers' API keys were exposed in logs. The company says they have been revoked.
- Private Chat: NEAR AI says there is no sign that stored conversations or uploaded files were accessed.
- Live requests: on 8 November the attacker could also have observed user requests being processed on the affected server, including confidential-mode requests. The report does not say how many.
- NEAR AI says NEAR Protocol and its other services were not affected.
What NEAR AI did and said
It revoked all active sessions and admin credentials, replaced all agent tokens, took Private Chat offline, deployed an SGLang fix across its inference servers and finished its credential rotations and a forensic review. It says it contacted affected users and agent owners directly and will make regulatory notifications where required. Its public guidance: if you have not heard from the company, no action is needed, and treat unsolicited messages that mention your account or balance with caution.
What to do
- If NEAR AI emailed you, follow the steps in that email and replace any API key you used on the platform.
- If you run SGLang yourself, look for the project's advisory and update when a fixed version is named. We could not find one. Meanwhile keep the inference server off the public internet and behind authentication. This is our advice, not the company's.
- Do not log API keys. Two keys were exposed through logs. Check what your own servers write to their logs.
- Be wary of messages that mention your NEAR AI account or balance and did not come from the company's address you already know.
- If none of this applies to you, NEAR AI says no action is needed.
What we could not confirm
We read NEAR AI's incident report and two press reports. We found no CVECVE A public ID number, like CVE-2026-12345, given to one specific security flaw so everyone can talk about the same bug. Tap again to open the glossary, no list of affected SGLang versions, no fixed version and no severity score in any of them, and the report does not link an SGLang advisory. No source says the flaw has been used against anyone else, but none rules it out. We did not read the company's posts on X. We do not know how many confidential-mode requests were observed or what the per-user emails say.
Watch the short video
The video loads from YouTube (privacy-enhanced mode) only when you press play.
Comments
No account needed. Add a nickname if you like, or leave it empty to comment anonymously. Comments are reviewed before they appear. Links are not allowed.
We store your comment, your nickname and a scrambled form of your connection that we use only to limit spam. We do not store your IP address or set cookies. Cloudflare Turnstile checks that you are not a bot.
Sources and verification
- NEAR AI, Security incident report: SGLang vulnerability (published 9 Nov 2026; read through a subagent) (opens in a new tab)
- The Block, NEAR AI says SGLang vulnerability exposed user account details in security breach (10 Nov 2026; read through a subagent) (opens in a new tab)
- crypto.news, NEAR AI fixes SGLang security flaw, restores Cloud service (10 Nov 2026; read through a subagent) (opens in a new tab)
We use primary sources first and say when we rely on a news report. Dates are the day we last checked each source.
Spotted a mistake? Email [email protected] and we will correct it and note the change.