The public needs exact access to the prompts and characteristics of internal models executing hacks to understand AI misalignment incidents

The public needs exact access to the prompts and characteristics of the internal models executing these hacks.

https://www.interconnects.ai/p/lessons-from-the-hacks