Was my NEAR AI Private Chat history stolen?

NEAR AI says it found no indication that stored Private Chat conversations or uploaded files were accessed, and it took Private Chat offline as part of its response. Requests being processed on the compromised inference server on November 8, 2026, including confidential-mode requests, could have been observed, so NEAR AI does not rule out exposure of data in transit that day.

Is there a CVE for the SGLang vulnerability NEAR AI reported?

No CVE appears in NEAR AI's incident report, and the company published no affected or fixed version numbers for SGLang. NEAR AI says it reported the vulnerability to the SGLang project and deployed a fix across all of its own inference servers. Operators running SGLang should follow the project's releases, because the report gives no version to pin to.

Did confidential computing fail to protect NEAR AI users?

Confidential computing worked as designed, according to NEAR AI, which explains that a trusted execution environment protects data from the host and the operator but not from a vulnerability in software running inside it. The remote code execution flaw sat in SGLang, the inference engine inside that environment. The 18 agents whose data was retrieved ran on a legacy deployment that was outside the environment entirely.

What did NEAR AI do after the breach?

NEAR AI revoked all active sessions and administrative credentials, replaced every agent token, took Private Chat offline, and deployed the SGLang fix across all inference servers. The company says it revoked the two exposed NEAR AI Cloud API keys, completed rotation of remaining credentials and a full forensic review, and notified users whose account details or agents were confirmed accessed.

HIGHIncident

NEAR AI: an SGLang flaw exposed 100 accounts and 18 agents

NEAR AI reports that an attacker used a remote code execution flaw in SGLang to reach user accounts and legacy agent data.

November 9, 2026 · NEAR AI · Research by NEAR AI Team

SGLang

HIGHIncident

NEAR AI says an attacker exploited a remote code execution flaw in SGLang, the inference engine on its servers, and took account details for about 100 users plus the chat history, settings and logs of 18 legacy agents.

01What happened

NEAR AI published a security incident report on November 9, 2026, after an attacker exploited a remote code execution vulnerability in SGLang, the open-source inference engine running on its servers. Account details for about 100 users were downloaded, along with the chat history, settings and logs of 18 agents on a legacy deployment. NEAR AI says it reported the vulnerability to the SGLang project.

02How it works

Remote code execution in the inference engine put the attacker inside the process that serves model requests. NEAR AI explains that confidential computing shields data from the host and the operator but not from a flaw in software running inside the enclave, which is what happened here. The 18 affected agents ran on a legacy deployment that sat outside the trusted execution environment altogether. Requests processed on the compromised server on November 8, 2026 could have been observed, including ones sent in confidential mode.

03Who is affected

About 100 NEAR AI user accounts had names and email addresses taken, and some also had plan and balance information downloaded. Eighteen agents on the legacy NEAR AI Agents deployment had their chat history, settings and logs retrieved. Two NEAR AI Cloud customers had API keys exposed in logs.

04What to do

Watch for direct contact from security@near.ai, which NEAR AI says carries the steps for each affected user and agent owner, and treat unsolicited messages that quote your account details or balance as suspect. Operators running SGLang themselves should follow the project's releases, because the report names no CVE and no version numbers.

Questions people ask

Was my NEAR AI Private Chat history stolen?: NEAR AI says it found no indication that stored Private Chat conversations or uploaded files were accessed, and it took Private Chat offline as part of its response. Requests being processed on the compromised inference server on November 8, 2026, including confidential-mode requests, could have been observed, so NEAR AI does not rule out exposure of data in transit that day.
Is there a CVE for the SGLang vulnerability NEAR AI reported?: No CVE appears in NEAR AI's incident report, and the company published no affected or fixed version numbers for SGLang. NEAR AI says it reported the vulnerability to the SGLang project and deployed a fix across all of its own inference servers. Operators running SGLang should follow the project's releases, because the report gives no version to pin to.
Did confidential computing fail to protect NEAR AI users?: Confidential computing worked as designed, according to NEAR AI, which explains that a trusted execution environment protects data from the host and the operator but not from a vulnerability in software running inside it. The remote code execution flaw sat in SGLang, the inference engine inside that environment. The 18 agents whose data was retrieved ran on a legacy deployment that was outside the environment entirely.
What did NEAR AI do after the breach?: NEAR AI revoked all active sessions and administrative credentials, replaced every agent token, took Private Chat offline, and deployed the SGLang fix across all inference servers. The company says it revoked the two exposed NEAR AI Cloud API keys, completed rotation of remaining credentials and a full forensic review, and notified users whose account details or agents were confirmed accessed.

Sources 2 links, all fetched

1near.aiNEAR AI Report
2docs.near.aiNEAR AI Verification