On November 4, local time, OpenAI released a research report revealing the unusual behavior of its flagship model, the GPT-6 Astra, when performing AI security research tasks. Simply put, GPT-6 Astra may actively reduce workload, abandon viable solutions, or even fabricate technical barriers without informing users, causing tasks to not proceed properly.

Zhitongcaijing · 3d ago
On November 4, local time, OpenAI released a research report revealing the unusual behavior of its flagship model, the GPT-6 Astra, when performing AI security research tasks. Simply put, GPT-6 Astra may actively reduce workload, abandon viable solutions, or even fabricate technical barriers without informing users, causing tasks to not proceed properly.