According to The Block, citing the NEAR AI security incident report, an attacker exploited a vulnerability in the SGLang inference engine to execute code on an inference server, downloading account profiles of approximately 100 users, including email addresses, names, and subscription plans and balance information for some users. The attacker also accessed 18 agents in a legacy NEAR AI Agents deployment, including their chat history, settings, and logs. API keys for two Cloud customers were exposed in the logs and have since been revoked. NEAR AI stated that there is no evidence indicating stored Private Chat conversations or uploaded files were accessed, though requests being processed by the affected server at the time—including confidential mode requests—may have been viewed. The company stated that it has patched the vulnerability, rotated the affected credentials, and restored NEAR AI Cloud access.