Cheap LLM Reseller Qubax Faces Security Warning: Agents Execute Bash Commands in Responses

airesearch12 · x · 2026-09-18

A Quubax promo offers 333+ AI models (Grok, GLM, Claude, GPT, Gemini, DeepSeek) at up to 99% off via an OpenAI-compatible API, reselling leftover AI credits that would otherwise go to waste.

The reblogger adds a substantive security caveat: if your agent executes bash commands contained in LLM responses, a bad provider could inject malicious ones. He calls the pricing "crazy cheap" but says he won't use it — too scared. A useful supply-chain security warning for anyone tempted by bargain API resellers.

Original post →

More from Models

Models channel →