Mohammad

One point worth noting: the OpenAI cyber models might have known those parts of your systems are exploitable because you had already shared your source code and used their models to code your environment. It might be the "insider knowledge".