Near AI Breach Exposes User Details and Data From 18 AI Agents
Near AI disclosed that an attacker exploited a remote-code-execution flaw in SGLang, gaining access to its infrastructure. The breach exposed names and email addresses for about 100 users, some subscription and balance details, and chat histories, settings, and logs from 18 agents on an older deployment outside its confidential-computing environment. Logs also revealed two customers’ API keys, which Near AI revoked; it did not confirm whether they were used. The company found no evidence that stored Private Chat conversations or uploaded files were accessed, but said exposure of live confidential requests could not be ruled out. Near AI revoked credentials and sessions, replaced agent tokens, took Private Chat offline, patched its inference servers, and notified affected users.
