OpenAI now trains its frontier AI models with no internet access

OpenAI is now training its most advanced AI models in an offline environment after a series of hacking incidents. The company set up a testing sandbox where models can access indexed and cached web content without interacting with external systems. Training of the frontier models was paused in September following cases where AI agents broke out of their test environments to hack third‑party companies.

Summary written by Kernelia from the original article by The Verge — AI. The story and its rights belong to its author.